Q PREDPOL

'
From Theory to Rractical Deployment
The Science Behind PredRol
Policing crime patterns is hard.
While crime may afflict the same neighborhoods year after year, the day-to-day
fluctuations in where and when crimes occur are large. Extensive research has shown that
day-to-day crime patterns are the result of: (1) crime generators that are fixed features of
the environment; (2) repeat and near-repeat victimization that leads previous victims and
their neighbors to be at greater risk of follow-on crimes; (3) the routine activity patterns of
offenders that keep risk local; and (4) substantial random noise. Each of these processes
is well known empirically, but when put together their impact on how crime hotspots
emerge, spread and disappear is incredibly complex. It is very hard to predict where crime
will occur in the next 10-12 hours given where it occurred yesterday.
Knowledge, skills and experience can reliably direct officers to the top two or three
riskiest locations in their operational environment. It is much harder for them to Identify
and choose between locations h e r e the risk may be lower and highly variable from day
to day. If an officer chooses to go to a location typically fourth in line for crime risk, but a
crime occurred at a location fifth in line on that day, then an opportunity to disrupt crime
is missed.
THE PREDICTIVE POLICING COMPANY."
The Rremise Behind
Predictive Policing
If one can accurately predict
where and when crimes will
occur; then law enforcement
personn'el can disrupt those
crimes before they happen.
Using high-powered mathematics and real-time crime data, PredPol evaluates yesterday's crimes in the context of all crimes occurring
over a long time horizon and wide spatial field to calculate accurate probabilities of where and when crime will occur today. Officers
using this information can make it harder for offenders to commit crimes in those locations leading to a net reduction in crime.
Predictive Rolicing Evidence from the Field
If predictive policing does not perform better than existing practice, or makes the job of the officer
on the street harder, then it is better to stick with existing practice.
Controlled experimental trials of PredPol were conducted with the Kent Police, UK, and Los Angeles Police Department, USA, with the
goal of assessing what (if any) advantages stem from the use of predictive policing. In each experiment, PredPol was tested head-to-
head against crime analysts using existing practice. PredPol predicts crime In 500' x 500' (150m x 150m) boxes with the number of
boxes deployed in a given shift calibrated to the policing resources available. The experimental deployments involved placement of 20
prediction boxes per shift in designated policing divisions. Analysts were tasked with deploying the same number of prediction boxes
per shift using all of the tools at their disposal. In Kent West Division, the analyst focused on intelligence-led policing practices, while Los
Angeles Foothill Division focused on crime hot spot mapping. In the absence of directed police patrol effects, PredPol predicts between
1.6-2.5 more crime than existing practice. Increased opportunities to impact crime are therefore of a similar magnitude using PredPol.
831.311.45501 infO@predpol.com 1 www.predpol.com PREDICT CRIME IN REAL TIME®
Q PREDPOL' THE PREDICTIVE POLICING
Accurate prediction leads not only to unique opportunities to disrupt crime, but also to serve the public at the scale at which crime
occurs. In live deployments In both Kent and Los Angeles, officers were directed to use available time between calls or appointments to
police what they see. Kent police officers capitalized on these opportunities:
06105113 Whilst potro!fing PredPo/ Zone 8 Harmer Street, Gravesend, Neighborhood officers attended a call to a suspicious mole
seen rooking into vehicfes. This male was seen to then break into a vehicle and due to the close proximity the officers they were
quickly on scene and detained the mole after a short foot chose. The male, a prolific offender, hod 2 Sot Navs in his pocket stolen
from 2 vehicles in Royal Pier rood. If officers hod not already been in the zone they would not have been in o position to catch the
offender as he committed the crime.
04106113 PredPof duty in Gravesend with 4 officers in 2 managed ro visit all PredPo/ boxes at feast once during the shrft with
different teams to show a high and varied presence, consisting of foot and mobile patrol. While checking vehicles they noticed 2
SAT NAV systems and money in the way of change on display, traced the registered owners of the vehicfes and gave them advice
over the phone, which was much appreciated.
02105113 From a local resident who has lived in the west side of Moidstone for many years: "great to see police around here again
you hove really made a big difference in cleaning up this port of town".
Disrupting crime during each and every shift can lead to significant crime declines over time. In Kent, the first six months of PredPol
deployment produced a steady decline In violent crime of -6%. In los Angeles, the first six months of deployment saw a -12% decline In
property crimes overall with a -25% decline in burglary alon.e.
The Limits of Predictive Policing
PredPo/ is about predicting where and when crime will occur; not who will commit a crime.
PredPol is not criminal profiling. It does not use any information about individuals or populations and their characteristics. The patterns
inherent In the crimes themselves provide ample information to predict where and when crimes will occur In the future.
Predictive policing disrupts the short-term, situational causes of crime. lt does not solve criminality, or the propensity for individuals
to commit crime. Predictive policing Is therefore not a replacement for policy and community engagement strategies needed to steer
people clear of criminal careers in the first place.
Key Contacts
PredPol- The Predictive Policing Company
Dr. P. jeffrey Brantingham
je/f@predpol.com
Los Angeles Police Department
Captain Sean Malinowski
sean.ma/inowski@lapd.lacity.org
Kent Police
Detective Chief Superintendent jon Sutton
jon.sutton@kent.pnn.police.uk
Mr. Mark Johnson, Head of Analysis
mark.johnson@kent.pnn.pofice.uk
831.331.4550 llnlo@predpol.rom I www.predpotrom PREDICT CRIME IN REAL TlMf®
@ 2013 PredPol, Inc. All dghls reserved. No part of this publication de sui bed herein may be reproduced, stored in a retrieval system, used in a spreadsheet, or transmitted
In any form or by means elettronlc, mechanical, photocopying, recording. or otherwise without the permission of PredPol,ln(,
From:
To:
Date:
Subject:
Donnie Fowler <donnie@predpol.com>
Susan.Giffin@sfgov.org, Rodrigo.Castillo@sfgov.org, greg.suhr@sfgov.org
Friday, September 06, 2013 10: 12AM
Completing the Predictive Policing Deployment
Chief Suhr, Susan, & Rod -
Page I of2
Happy September to you. I write to let you see the SFPD predictive policing system as it will
run when you are ready to deploy. It's one thing when I see deployments in another city and
something else entirely when I see predictions where I live here in S. F.
You can find it with the simple log-in name and password below:
* once you log in, click on
district
red box at the top right and select the whole city or any
Per our most recent conversation, we were waiting for the Oracle work to finalize before
finishing our integration with you on general property crimes and on the first-in-the-r;Jation
rollout of gun violence predictions·(along with Seattle and Atlanta). If my memory is correct, you
thought September would be the likely time.
In the last few months, we have had tremendous success with deployments from Alhambra, CA,
to Norcross, GA, to Kent, England. You will see some of our work in the attached white papers
on results and on gun violence. There are also a few news stories below my signature.
All the best,
Donnie
DONNIE FOWLER Director of Business Development
.. 4th Street, ..I
San Francisco, CA 94114
C : 4 1 · · · ~
PREDPOL.COM
The Predictive Policing Company
In the News ...
Fox News· Atlanta, August 2013 ("On the first day, they apprehended a suspect in the middle of committing a
burglary.")
http://www.myfoxatlanta.com/story/23177698/computer-tries-to-predict
The Economist, July 2013 ("easier to foresee wrongdoing and spot likely wrongdoers")
http://www.economist.com/news/briefinq/21582042-it-gettinq-easier-foresee-wrongdoinq-and-spot-likely-
http://sfmail04.sfgov.org/mail/gsuhr.nsf/(%24Inbox)/8C34B97B743B991 D6276BA51 EE321... 9/23/13
Page 2 of2
wrongdoers-dent-even-think-about-it
NBC News- Los Angeles, January 2013 ("a cliff-like drop when predictive policing began")
http://www.nbclosangeles.com/news/locai/LAPD-Chief-Charlie-Beck-Predictive-Policinq-Forecasts-Crime-185970452.html
Current TV with Santa Cruz Crime Analyst & Fmr. Michigan Gov. Jennifer Granholm, January 2013
("Science fiction has become science fact.")
https://current.box.com/s/c6bxouhsdugp3ifngz4s
San Francisco Chronicle, August 2012 ("a new web-bases system to monitor real· time crime data and predict
when the next violence will occur")
http://www.sfgate.com/bayarea/article/SF-mayor-announces-antiviolence-strategy-3770653.php
CBS Evening News - National, March 2012 ("It's not how many people you catch, it's how many crimes you
prevent")
404996n
Attachments:
White Paper Predicting Gun Violence (2013
July).pdf
White Paper Science and Testing of Predictive
Policing (2013).pdf
http://sfmail04.sfgov.org/mail/gsuhr.nsf/(%241nbox)/8C34B97B743B991D6276BA5l EE32l... 9/23/13
Page I of3
From: Suzy Loftus <suzyloftus@hotmail.com>
To: Donnie Fowler <dfowler@gmail.com>, "greg.suhr@sfgov.org" <greg.suhr@sfgov.org>
Date: Thursday, May 09, 2013 01:42PM
Subject:
RE: Predictive Policing & Gun Violence
Thanks, Donnie.
Chief-- just let me know how you want to proceed. This is incredibly exciting and I think there
would be significant public Interest in the department utilizing this technology. Let me know how
I can be helpful.
Suzy
Date: Thu, 9 May 2013 11:48:17-0700
Subject: Predictive Policing & Gun Violence
From: dfowler@gmail.com
To: greg.suhr@sfgov.org
CC: suzyloftus@hotmail.com
Chief Suhr-
Thanks again for your decision to deploy predict'ive policing in San Francisco. We are just about
ready to roll, so I write to give you an update specifically on the new, first-In-the-nation gun
violence prediction technology that you will soon have as part of the overall PredPol tool. You
probably know that SFPD will be one of the first three cities, with Atlanta and Detroit. to use
predictive analytics to predict gun violence.
We hope you and the Mayor will agree. please. to join the announcement with those two cities on
May 23 or 24. This could be as significant as the kind of press conference the Seattle mayor and
chief had a couple of months ago (see below) or as simple as a joint press statement that San
Francisco will once again show up as a leader on tech innovation.
With Shotspotter locating gunshots after they happen, PredPol will be able to predict where some
of this violence might occur before It happens. We are glad to be working on a partnership with
them to make both of these tactics better. And we are glad to be working with both Susan Giffin
and Rod Castillo in your office to implement things.
Please do not hesitate to contact me with any suggestions or questions.
All the best,
Donnie Fowler
San Francisco
predpol. com
415- cell
attachment: gun violence white paper
Predictive Policing in the News ...
http://sfmail04.sfgov .org/mail/gsuhr.nsf/(%24 Inbox)/3 D E64C I F76A2E6F3 8E39129ED8A9... 9/23/13
Page 2 of3
Seattle Times, February 2013 ("will allow us to be proactive rather than reactive")
http:i/bloqs.seattletimes.com/today/2013/02/seattle-police-turn-to-computer-software-to-fiqht-crime/
KPLU-Seattle (NPR), February 2013 ("will help them allocate patrols more effectively")
http://www.kplu.org/posUseattle-tacoma-rollinq-out-new-predictive-policing-software
NBC News -Los Angeles, January 2013 ("a cliff-like drop when predictive policing began")
http://www.nbclosangeles.com/news/locai/LAPD-Chief-Charlie-Beck-Predictive-Policing-Forecasts-Crime-
185970452.html
Current TV with Santa Cruz Crime Analyst, January 2013 ("science fiction has become science fact")
https://current.liox.com/s/c6bxouhsdugp3ifngz4s [with former Michigan Governor Jennifer Granholm]
On Fri, May 18, 2012 at 3:59PM, Donnie Fowler <dfowler@qmail.com> wrote:
Chief Suhr-
Suzy Loftus was good enough to reconnect us, so I'm following up with some details
on the predictive policing technology she mentioned to you. A summary is attached
to this email that you can share with your team. We will send along some additional
technology details shortly and look forward to talking.
We are moving from successful field experiments to full deployment this summer. It
would be great to see San Francisco (where I also live) take a front-row seat in the
state and nationally on this.
All the best,
Donnie Fowler
~ c e l l
Predictive Policing
The Problem: Police departments nationwide are facing budget freezes and deep cuts, requiring
them to manage their resources more effectively while still responding to public demand for crime
prevention and reduction.
The Solution: Like forecasting the weather, PredPol's patent-pending technology generates
predictions about which areas and windows of time are at highest risk for future crimes. In
http://sfmai 104.sfgov .org/mail/gsuhr.nsf/(%24 Inbox)/3DE64C I F76A2E6F38E39129ED8A9 ... 9/23/13
Page 3 of 3
contrast to analysis that simply maps past crime data, this technology applies advanced
mathematics and computer learning that has resulted in predictions twice as accurate as those
made by experienced crime analysts and veteran police using only their own knowledge and
experience.
---"------ Forwarded message ----------
From: Suzy Loftus <suzyloftus@hotmail.com>
Date: Fri, May 18, 2012 at 3:28PM
Subject: Re: Predictive Policing and all things good law enforcement
To: Donnie Fowler <dfowler@qmail.com>
Cc: Debbie Mesloh <debmesloh@yahoo.com>, Caleb Baskin
<csbaskln@baskinqrant.com>
Hey guys,
Good to meet both of you today - I am fascinated by what is possible here. I called
Chief Suhr about it and told him that I met with you guys and think if he likes it, it
could be great for sf. He would like you two to send him an email and he'll get his
technology person lined up to do an initial vetting on how it would work and then .
he'll follow up. His email is Greq.suhr@sfqov.org.
Let me know how It goes.
Best,
Suzy
Donnie Fowler
4 1 · - ~
dfowler@qmail.com
http://sfmai104.sfgov .org/mail/gsuhr.nsf/(%24Inbox)/3DE64C 1 F76A2E6F38E3 9129ED8A9... 9/23/13
From:
To:
Cc:
Date:
Subject:
Donnie Fowler <dfowler@gmail.com >
greg.suhr@sfgov .org
Suzy Loftus <suzyloftus@hotmail.com>
Thursday, May 09, 2013 11:47AM
Predictive Policing & Gun Violence
Chief Suhr-
Page I of3
Thanks again for your decision to deploy predictive policing in San Francisco. We are just about
ready to roll, so I write to give you an update specifically on the new, first-in-the-nation gun
violence prediction technology that you will soon have as part of the overall PredPol tool. You
probably know that SFPD will be one of the first three cities. with Atlanta and Detroit. to use
predictive analytics to predict gun violence.
We hope you and the Mayor will agree. please. to join the announcement with those two cities on
May 23 or 24. This could be as significant as the kind of press conference the Seattle mayor and
chief had a couple of months ago (see below) or as simple as a joint press statement that San
Francisco will once again show up as a leader on tech innovation.
With Shotspotter locating gunshots after they happen, PredPol will be able to predict where some
of this violence might occur before it happens. We are glad to be working on a partnership with
them to make both of these tactics better. And we a ~ : e glad to be working with both Susan Giffin
and Rod Castillo in your office to implement things.
Please do not hesitate to contact me with any suggestions or questions.
All the best,
Donnie Fowler
ell
attachment: gun violence white paper
Predictive Policing in the News ...
Seattle Times, February 2013 ("will allow us to be proactive rather than reactive")
http://blogs. seattletimes. com/today/2013/02/seattle-police-turn-to-computer -software-to-fight -crime/
KPLU-Seattle (NPR), February 2013 ("will help them allocate patrols more effectively")
http://www. kplu .org/posUseattle-tacoma-rolling-out -new-predictive-policing-software
NBC News - Los Angeles, January 2013 ("a cliff-like drop when predictive policing began")
http://sfmail04.sfgov .org/mail/gsuhr.nsf/(%24 Inbox)/4 FBE67F 52A4 0561 E89 A264 B89E5 F... 9/23/13
Page 2 of3
http://www. nbclosanqeles. com/news/local/LAP D-Ch ief -Charlie-Beck -Predictive-Policinq-F orecasts-Crime-
185970452.html
Current TV with Santa Cruz Crime Analyst, January 2013 ("science fiction has become science fact")
https://current.box.com/s/c6bxouhsduqp3ifngz4s (with former Michigan Governor Jennifer Granholm]
On Fri, May 18, 2012 at 3:59PM, Donnie Fowler <dfowler@gmail.com> wrote:
Chief Suhr-
Suzy Loftus was good enough to reconnect us, so I'm following up with some details on the
predictive policing technology she mentioned to you. A summary is attached to this email that
you can share with your team. We will send along some additional technology details shortly
and look forward to talking.
We are moving from successful field experiments to full deployment this summer. It would be
great to see San· Francisco (where I also live) take a front-row seat in the state and nationally
on this.
All the best,
Donnie Fowler
415 cell
Predictive Policing
The Problem: Police departments nationwide are facing budget freezes and deep
cuts, requiring them to manage their resources more effectively while still
responding to public demand for crime prevention and reduction.
The Solution: Like forecasting the weather, PredPol's patent-pending technology
generates predictions about which areas and windows of time are at highest risk
for future crimes. In contrast to analysis that simply maps past crime data, this
technology applies advanced mathematics and computer learning that has
resulted in predictions twice as accurate as those made by experienced crime
analysts and veteran police using only their own knowledge and experience.
---------- Forwarded message ----------
From: Suzy Loftus <suzyloftus@hotmail.com>
Date: Fri, May 18, 2012 at 3:28PM
Subject: Re: Predictive Policing and all things good law enforcement
To: Donnie Fowler <dfowler@gmail.com>
Cc: Debbie Mesloh <debmesloh@yahoo.com>, Caleb Baskin <csbaskin@baskingrant.com>
Hey guys,
http://sfmail04.sfgov .org/mail/gsuhr.nsf/(%24lnbox)/4 FBE67F 52A4D561 E89 A264 B89E5F ... 9/23/13
Page 3 of3
Good to meet both of you today- I am fascinated by what is possible here. I called Chief Suhr
about it and told him that I met with you guys and think if he likes it, it could be great for sf.
He would like you two to send him an email and he'll get his technology person lined up to do
an initial vetting on how it would work and then he'll follow up. His email is
Greq.suhr@sfgov .org.
Let me know how it goes.
Best,
Suzy
Donnie Fowler
415 c
dfowler@qmail.com
Attachments:
Predicting Gun Violence White Paper (2013 April).pdf
http://sfmail04.sfgov .org/mail/gsuhr.nsf/(%24 Inbox)/4 FBE67F52A4 0561 E89 A264B89E5F... 9/23/13
Page I of 4
From: Susan Giffin/SFPD/SFGOV
To: Greg Suhr/SFPD/SFGOV@SFGOV, David.Shinn@flysfo.com
Date: Monday, May 13, 2013 12: 14PM
Subject:
Fw: Predicting Gun Violence in SF, ATL, and Detroit May 23-24
History:
<··This message has been replied to and forwarded.
Chief-
I corresponded with PredPol and suggested that we not participate in this announcement at this
time (prior to message below). While we will be rolling out PredPol and the gun violence module,
I would like to wait until we are fully implemented before any announcements. I can elaborate
further in person if you would like. Thanks.
Susan
Susan Giffin
Chief Information Officer, SFPD
(415) 553-1481 (Angel Yee)
------Forwarded by Susan Giffin/SFPD/SFGOV on 05/13/13 12:13 -----
From: Greg Suhr/SFPD/SFGOV
To: "Susan Giffin" <susan.giffin@sfgov.org>, "Lyn Tomioka" <Lyn.Tomioka@sfgov.org>
Date: 05/09/13 18:05
Subject: Fw: Predicting Gun Violence In SF, ATL, and Detroit May 23-24
FYI
From: "Winnicker, Tony" [tony.winnicker@sfgov.org]
Sent: 05/10/2013 12:50 AM GMT
To: Christine Falvey; Jason Elliott
Cc: Greg Suhr
Subject: FW: Predicting Gun Violence in SF, ATL, and Detroit May 23-24
See below- Chief- are we working to deploy this technology? See below.
Christine, Jason- they would like us to join/consider an announcement with A TL and Detroit.
Donnie Fowler spoke to the Mayor about this following the Sandy Hook Promise gun event last
month at Bill Graham.
From: Donnie Fowler [mailto:dfowler@qmail.com]
Sent: Thursday, May 09, 2013 5:42 PM
http://sfmail04 .sfgov .org/mail/gsuhr.nsf/(%24 Inbox)/C32AB489504 7206688257B6A00697 A... 9/23/13
Page 2 of 4
To: Winnicker, Tony
Subject: Predicting Gun Violence in SF, ATL, and Detroit May 23-24
Tony-
Good to talk with you this afternoon about San Francisco's decision to be the first to deploy
technology that predicts gun violence along with Atlanta and Detroit. We at PredPol, based here in the
Bay Area, have had a great working relationship with ChiefSuhr and SFPD's CIO Susan Giffin as
they integrate predictive analytics into their overall technology plan. We are also partnering with the
Sandy Hook Promise group.
As I mentioned, Atlanta's Mayor and Chief want to announce that they intend to implement this gun
violence methodology on May 23 or 24th. We hope the Mayor and Chief will agree. please, to join the
announcement with the other two cities on May 23 or 24. This could be as significant as the kind of press
conference the Seattle mayor and chief had a couple of months ago (see below) or as simple as a joint press
statement that San Francisco will once again show up as a leader on tech innovation.
Please do not hesitate to contact me with any suggestions or questions.
All the best,
Donnie Fowler
San Francisco
c I
attachment & copied below: gun violence summary points
Predictive Policing in the News ...
Seattlq Times, February 2013 ("will allow us to be proactive rather than reactive")
http :1/blogs. seattleti mes. com/tod a y/2 0 13/02/seattle-pol ice-turn -to-com puler-software-to-fight-
crime/
KPLU-Seattle (NPR), February 2013 ("will help them allocate patrols more effectively")
h tip: I /www. kpl u. org/pos Useattle-tacoma-rolling-out -n ew-p red ictive-po !icing-software
NBC News- Los Angeles, January 2013 ("a cliff-like drop when predictive policing began")
http://www.nbclosangeles.com/news/locai/LAPD-Chief-Charlie-Beck-Predictive-Policing-
Forecasts-Crime-185970452.html
Current TV with Santa Cruz Crime Analyst, January 2013 ("science fiction has become
science fact")
https://current.box.com/s/c6bxouhsdugp3ifngz4s [with former Michigan Governor Jennifer
Granholm]
DRAFT[May8, 2013]
PREDICTING GUN VIOLENCE
Cities Take Lead as First in the Nation to Deploy New Technology to Deter Gun Cl'ime
http:/ /sfmail04.sfgov .org/mai 1/gsuhr.nsf/(%24 Inbox)/C32AB489504 7206688257B6A00697 A... 9/23/13
Page 3 of 4
The Challenge. In light of recent gun violence across the country, including school
shootings, and as police departments across the nation face tighter budgets and scarcer
resources, reducing violence and gun related crime remains a significant challenge.
The Leaders. Atlanta, Detroit, and San Francisco are taking the lead as the first cities to
predict and deter gun violence by deploying a new application of PredPol's general crime
prediction methodology.
Predictive Policing. While no one strategy serves as a silver bullet, predictive policing
gives officers a significantly better idea of when and where to be so that they can deter crime
generally and gun violence in particular, sometimes even catching criminals in the act.
Precursor Crimes. PredPol's unique gun violence prediction methodology recognizes that
past homicides are not necessarily the best predictors of future gun violence. In fact, the
occurrence of serious violent crimes like weapons violations, assault, and battery provide as
much, or more, information on where and when future homicides are most likely to occur.
Origins & Results of Predictive Policing. Predictive policing was first developed and
deployed in Los Angeles and Santa Cruz, California. Only six months after launch, those two
cities enjoyed declines ranging from -12% to -25% in burglaries, car thefts, and thefts from
motor vehicles compared to the same periodjn the previous year.
Field Tested. PredPol's general crime prediction technology has been extensively evaluated
using historical crime data and controlled field trials with multiple law enforcement agencies
over several years. Controlled trials show that PredPol predicts more than two-times as much
crime as a veteran police and crime analysts. To be clear, though, PredPol is not a
replacement for veteran officers and crime analysts.
Modeled Using Real Crime Data. PredPol has also extensively modeled its new gun
violence methodology, predicting a greater number of gun homicides compared with
alternative approaches, including traditional hotspot maps. In a test of public crime data out
of Chicago, PredPol successfully predicted so% of gun homicides by flagging only 10.3% of
that city-- 66% more than hotspot mapping would predict.
Science for Patrol Officers. Developments in mathematical and statistical modeling, high-
performance cloud computing, and GPS-enabled mobile devices now make it possible to
deliver real-time crime forecasts on simple maps to patrol officers.
Not Crime Mapping. Without predictive analytics, police end up chasing yesterday's crime
by relying only on intuition and mapping of past crime data. As LAPD's Captain Sean
Malinowski has noted, "We look at these maps and they're as accurate as we can get them.
But I'm looking at a map from last week and the whole assumption is that next week is like
last week." Traditional mapping tools are calibrated less frequently, rely more on humans to
recognize patterns, and allocate resources based on past crimes rather than predicted future
offenses.
Timing. Although the actual deployment date for Atlanta, Detroit, and San Francisco differ,
http :II sfma i I 04 .sfgov .org/ma iII gsuhr.nsf/(%24lnbox )/C3 2A B48 9 504 720668825 7 B6A 00697 A... 9/23/13
Page 4 of 4
these cities have smartly added a new tool in the expanding national efforts to reduce gun
violence.
(See attached file: Gun Violence Announcement Summary Points PredPol (2013-05-0B).docx)
Attachments:
Gun Violence Announcement Summary Points PredPol (2013-05-0B).docx
http:// sfmail 04. sfgov .org/mail/ gsuhr .nsf/ (%24 In box )/C3 2AB4 8 9 504 72066 8 825 7 B6A00697 A... 9/23/13
Page l of l
From: Susan Giffin/SFPD/SFGOV
To: Suzy Loftus <suzyloftus@hotmail.com>
Cc: "greg.suhr@sfgov.org" <greg.suhr@sfgov.org>, Rodrigo Castillo/SFPD/SFGOV@SFGOV
Date:
Subject:
Tuesday, May 28, 2013 !0:09AM
Re: Article about predictive policing
Hi Commissioner Loftus. Thank you for the article -good information. We have already implemented a pilot
version of this software. We can actually see our hot spots (our IT group can see them). However, we still have
work to do to integrate the PredPol software and test it fully and we need Oracle for that. The hold up has been
that we had to stop work with Oracle in order to renew our contract - and the city's contract renewal has taken
over 6 months.
Short answer is I don't have a date yet and it is not likely to be before summer (which is upon us.) I will let you
know as soon as we get this into production. Thanks as always for your support.
Susan
Susan Giffin
Chief Information Officer, SFPD
(415) 553-1481 (Angel Yee)
From· Suzy Loftus <suzytoftus@hotmail.corri>
To: "susan.giffin@sfgov.org" <susan.giffin@sfgov.org>, "greg.suhr@sfgov.org" <greg.suhr@sfgov.org>
Date: 05/28113 09:37
Subject: Article about predictive policing
Chief Suhr and Director Giffin,
I found the attached article in the Santa Clara University magazine-- it describes the methodology
and the team that built the algorithm behind predictive policing. They also reference a number of
other jurisdictions that are using the technology and infusing other data to achieve different goals. It
is very exciting. Please keep me posted on our progress rolling this out -- is the plan still to use the
technology to predict and deploy patrols to suppress gun violence? Are you going to deploy it
before the summer when kids are out?
I'm sure it is still in the early stages, but just wanted to get an update. Thanks, in advance.
Suzy
Attachments:
Article on Predictive Policing SCU Magazine.pdf
http:/ /sfmail04.sfgov .org/mail/gsuhr.nsf/(%24 Inbox)/77EED E4875A l3E7088257B79005DA... 9/23/13
From:
To:
Date:
Donnie Fowler <donnie@predpol.com>
Donnie Fowler <donnie@predpol.com>
Thursday, July 25, 2013 07:23AM
Subject: The Economist: "Predictive Policing -- Don't Even Think About It"
The Economist
Jul 20th 20131From the print edition
Predictive Policing
Don't Even Think About It
It is getting easier to foresee wrongdoing and spot likely wrongdoers
http://www. economist. com/news/briefing/2158 2 04 2-it -getting-easier-foresee-wrongdoing-and-spot -likely-
wrongdoers-doni-even-think-about-it
Page I of3
THE meanest streets of Kent [England] are to be found in little pink boxes. Or at least they are if you look at them
through the crime-prediction software produced by an American company called PredPol. Places in the county
easl of London where a crime is likely on a given day show up on PredPol's maps highlighted by pink squares 150
metres on a side. The predictions can be eerily good, according to M.ark Johnson, a police analyst: "In the first box
I visited we found a carving knife just lying in the road."
Pre.dPol is one of a range of tools using better data, more finely crunched, to predict crime. They seem to promise
better law-enforcement. But they also bring worries about privacy, and of justice systems run by machines not
people.
Criminal offences, like infectious disease, form patterns in time and space. A burglary in a placid neighbourhood
represents a heightened risk to surrounding properties; the threat shrinks swiftly if no further offences take place.
These patterns have spawned a handful of predictive products which seem to offer real insight. During a four-
month trial in Kent, 8.5% of all street crime occurred within PredPol's pink boxes, with plenty more next door to
them; predictions from police analysts scored only 5%. An earlier trial in Los Angeles saw the machine score 6%
compared with human analysts' 3%.
Intelligent policing can convert these modest gains into significant reductions in crime. Cops working with
prepictive systems respond to call-outs as usual, but when they are free they return to the spots which the
computer suggests. Officers may talk to locals or report problems, like broken lights or unsecured properties, that
could encourage crime. Within six months of introducing predictive techniques in the Foothill area of Los Angeles,
in late 2011, property crimes had fallen 12% compared with the previous year; in neighbouring districts they rose
0.5% (see chart). Police in Trafford, a suburb of Manchester in north-west England, say relatively simple and
sometimes cost-free techniques, including routing police driving instructors through high-risk areas, helped them
cut burglaries 26.6% in the year to May 2011, compared with a decline of 9.8% in the rest of the city.
For now, the predictive approach works best against burglary and thefts of vehicles or their contents. These
common crimes provide plenty of historical data to chew on. But adding extra types of information, such as details
of road networks, can fine-tune forecasts further. Offenders like places where vulnerable targets are simple to spot,
access is easy and getaways speedy, says Shane Johnson, a criminologist at University College London. Systems
devised by IBM, a technology firm, watch how big local events, proximity to payday and the weather affect the
frequency and location of lawbreaking. "Muggers don't like getting wet," says Ron Fellows, IBM's expert. Jeff
Brantingham of PredPol thinks that finding speedy ways to ingest crime reports is more important than adding data
sets. Timelier updates would allow PredPol to whirr out crime predictions constantly, rather than once per shift. Mr
Fellows enthuses about sensors that detect gunshots (already installed in several American cities) and smart
CCTV cameras that recognise when those in their gaze are acting suspiciously. He promises squad cars directed
by computers, not just control centres, which could continually calculate the most useful patrol routes.
http://sfmail04. sfgov.org/ mail/gsuhr.nsf/(%24Inbox)/64A6E214CAF03B6ADC2CE41 09847... 9/23/ 13
Page 2 of3
Minority report
Predicting and forestalling crime does not solve its root causes. Positioning police in hotspots discourages
opportunistic wrongdoing, but may encourage other criminals to move to less likely areas. And while data-
crunching may make it easier to identify high-risk offenders-about half of American states use some form of
statistical analysis to decide when to parole prisoners-there is little that it can do to change their motivation.
Misuse and overuse of data can amplify biases. It matters, for example, whether software crunches reports of
crimes or arrests; if the latter, police activity risks creating a vicious circle. And report-based systems may favour
rich neighbourhoods which turn to the police more readily rather than poor ones where crime is rife. Crimes such
as burglary and car theft are more consistently reported than drug dealing or gang-related violence.
But mathematical models might make policing more equitable by curbing prejudice. A suspicious individual's
presence in a "high-crime area" is among the criteria American police may use to determine whether a search is
acceptable: a more rigorous definition of those locations will stop that justification being abused. Detailed analysis
of a convict's personal history may be a fairer reason to refuse parole than similarity to a stereotype.
Technology may also sharpen debates about what people want from their justice systems, and what costs they are
willing to accept. For example, software developed by Richard Berk, an American statistician, which is credited
with helping to cut recidivism among paroled prisoners in Philadelphia, requires the authorities to define in advance
their willingness to risk being overly tough on low-risk offenders or to under-supervise nasty ones.
This sort of transparency about what goes on in predictive systems, and what their assumptions are, may also be a
partial solution to worries voiced by Andrew Ferguson, a law professor in Washington, DC. Mr Ferguson fears that
judges and juries could come to place too much credence in the accuracy of crime prediction tools, jeopardising
justice. If transparency is a good counter to this, it will be important to preserve it as prediction becomes a bigger
business and gets further from its academic roots. ·
It is as prediction moves from places to people that it becomes most vexed. Police attending domestic
disturbances in Los Angeles have tried out a checklist, derived from much data-crunching, to determine whether
the incident presages violence. Mr Berk is working with authorities in Maryland to predict which of the families
known to social services are likely to inflict the worst abuses on their children. Federal officials aim to forecast
potential health and safety infringements. America's Department of Homeland Security is seeking to perfect
software which scans crowds or airport queues to detect nervous behaviour such as fidgeting, shallow breathing
and signs of a swift heartbeat.
So far, predictions have mostly been made about people who have already had contact with the justice system-
such as convicted criminals. The growth of social media provides a lot of crunchable data on everyone else. Firms
that once specialised in helping executives measure how web users feel about their brands now supply products
that warn police when civil unrest approaches, and help them closely follow crises. Cops in Cillifornia admit to
trawling social networks for early warnings of wild parties. ECM Universe, an American firm, offers software that
crawls sites "rife with extremism" to identify people who deserve closer attention.
The legal limits on using social media to fish out likely wrongdoers, or create files on them, are contested. Most
laws governing police investigations pre-date social networking, and some forces assert that all information posted
to public forums is fair game. But Jamie Bartlett of Demos, a British think-tank, says citizens and police forces
need clearer guidance about how to map physical-world privacy rights onto online spaces. He thinks gathering
information about how someone behaves on social sites ought to require the same clearance needed to monitor
them doggedly in public places. Officers who register anonymously or pseudonymously to read content, or send
web crawlers to trawl sites against their owner's wishes, would require yet more supervision.
Identifying true villains among the oddballs and loudmouths found by social-media searches is tricky. Most police
efforts are embryonic. Evgeny Morozov, an academic and technology writer, thinks the privacy-conscious have
more to fear from crime detection algorithms cooked up by social networks themselves. Some of those firms
already alert investigators when they suspect users of soliciting minors. Unlike the cops they employ clever coders
who can process private messages and other data that police may access only with a court order.
These projects make life difficult for many criminals. But smart ones use the internet to make predictions of their
http://sfmai 104.sfgov .org/mail/gsuhr.nsf/(%241nbox)/64A6E214CAF03B6ADC2CE41 0984 7... 9/23/13
Page 3 of 3
own. Nearly 80% of previously arrested burglars surveyed in 2011 by Friedland, a security firm, said information
drawn from social media helps thieves plan coups. Status updates and photographs generate handy lists of
tempting properties with absent owners. It does not take a crystal ball to work out what comes next.
From the print edition: Briefing
DONNIE FOWLER Director of Business Development
331 Soquel Ave. Ste. 100
Santa Cruz, CA 95062
15
PREDPOLCOM
The Predictive Policing Company
http://sfmail04.sfgov .org/mail/gsuhr.nsf/(%24 Inbox)/64A6E214CAF03B6ADC2CE41 0984 7... 9/23/13
Pagel of3
From: Donnie Fowler <donnie@predpol.com>
To: Donnie Fowler <donnie@predpol.com>
Date: Thursday, July 25, 2013 07:22AM
Subject: Predicting Gun Violence: PredPol Deploys First-Of-Its-Kind Technology
First-Of-Its-Kind Technology Predicts Gun
Violence,
Providing Another Answer to Gun Crime
http://www .predpol.com /gun-violence I
also see attached white paper on gun violence predictions
Santa Cruz I Silicon Valley, California -- PredPol, the Predictive Policing Company, has
developed and is deploying first-of-its-kind technology to help police predict and deter
gun violence. · ·
"We are pleased to be working with innovative leaders in our partner cities," said Dr.
George Mohler, PredPol co-founder and professor of mathematics and computer
science at Santa Clara University. "While no single strategy can end gun violence,
predictive policing gives officers a significantly better idea of when and where to be so
that they can deter crime generally, and gun violence in particular. If we can prevent
just one gun crime, that will mean one less neighbor, friend, or family member who
becomes a victim."
PredPol's unique gun violence prediction methodology leverages existing crime data that
every city already has, advanced mathematics developed over more than six years,
computer learning, cloud computing, and the indispensable experience of veteran
police. The crime data is analyzed through a sophisticated algorithm that applies proven
criminal theories about crime in general and gun violence in particular. The results are
more accurate and more actionable recommendations for when and where gun violence
is most likely to occur, thus allowing police to show up before crime happens.
One key element of the gun violence methodology uses so-called precursor crimes to
best determine where and when future violence will happen. Specifically, and perhaps
http:/ I sfmail 04 .sfgov .org/mail/ gs uhr.nsf/(%24 In box )/25 8 862 868CFD972 90 36AF CD6A E02... 9/23/ 13
Page 2 of3
surprisingly, past homicides are not necessarily the best predictors of future homicides.
The occurrence of incidents like weapons violations, assaults, and batteries provide as
much or more information about where and when future homicides are most likely to
occur. In a manner similar to 'broken-windows' policing, targeting these precursor
crimes can have a significant impact on gun homicide and ultimately on crime reduction.
PredPol has extensively modeled this new approach to deterring gun violence, predicting
a greater number of gun homicides compared with existing approaches, including
traditional hotspot maps. PredPol successfully outperforms current best practices by
66%.
This corresponds with even stronger performance that police agencies have seen with
PredPol's broader technology for predicting and preventing other violent crimes and
property offenses. In deployments to patrol officers in cities like Los Angeles and Kent,
England, PredPol's predictions not only· reduced crime but were 100% more accurate
than existing best practices - predicting double the number of crimes as current crime
mapping and hotspot analysis.
PredPol has already deployed its broad crime prediction tool for less violent crimes and
for property offenses in cities like Los Angeles and Santa Cruz in California, South
Carolina's capital city of Columbia, Seattle and Tacoma in Washington, and in Kent,
England. The gun violence prediction technology is part of this broader predictive
policing technology package.
"We understand the challenges faced by city leaders and law enforcement and created
this technology with police to help them do more with less," said Dr. Jeff
Brantingham, co-founder of PredPol and a criminology expert at UCLA. "Affordable,
easy to use technologies allow police who have tight budgets and limited hiring ability to
better direct the patrol resources they have."
Without predictive analytics, police are forced to chase yesterday's crime by relying on
simple mapping of past crime data. It would be like forecasting the weather using only
historical weather patterns, but ignoring weather radar. Traditional crime mapping tools
are calibrated less frequently, rely more on humans to recognize patterns, and allocate
resources based on past crimes rather than predicted future offenses. PredPol does not
replace the insights of veteran officers and crime analysts, but delivers an easy-to-use
enhancement that lets police do more with their current resources.
http:// sfmail 04. sfgov .org/maill gsu hr. nsf/ (%24 Inbox)/25 8B62868CF 0972 9 D 3 6AF CD6AE02... 9/23/13
Page 3 of3
About PredPol
A secure, cloud-based software-as-a-service, PredPol's tool was developed by a team of
PhD mathematicians, criminologists, and social scientists at UCLA, Santa Clara
University, and UC Irvine in close collaboration with crime analysts and line level officers
at the Los Angeles and Santa Cruz Police Departments. Just six months after launch,
those first two cities saw crime reductions of 12% to 25% in burglaries and auto thefts
compared to the previous year.
After successful initial deployments in 2011 in Los Angeles and Santa Cruz, PredPol has
deployed its groundbreaking technology in dozens of cities around the United States as
well as the United Kingdom. PredPol's core technology has grown from success with
property crimes to include prediction of drug crime, gang crime, anti-social behavior,
and now .gun violence.
For more information please visit < http://www.predpol.com/gun-violence > and see
the attached white paper.
DONNIE FOWLER Director of Business Development
331 Soquel Ave. Ste. 100
Santa Cruz, CA 95062
PREDPOL.COM
The Predictive Policing Company
Attachments:
White Paper Predicting Gun Violence (2013
July).pdf
55 Map with Predictive Boxes.JPG
http:/ /sfmai I 04 .sfgov .org/ma i If gsuhr. nsf/(%24 In box )/25 8B62 868 CFD972 9 D 3 6AF CD6AE02... 9/23/ 13
From:
To:
Date:
Subject:
Donnie Fowler <donnie@predpol.com>
Donnie Fowler <donnie@predpol.com>
Friday, July 26, 2013 02:31PM
NPR: Predicting Gun Violence ("All Things Considered," Friday, July 26)
Can Software That Predicts Crime Pass Constitutional Muster?
by MARTIN KASTE, NATIONAL PUBLIC RADIO
"All Things Considered," Friday, July 26, 2013
Page I of 4
It may be that PredPol is a constitutional basis for stopping someone. Some might consider it more ol
police officer's judgment- less prone to racism or other kinds of profiling, for example.
http://www. n pr. orq/20 13/0 7/2 6/205 83 56 7 4 I can-software- that- predicts-crime- pass- constitul
Typically, police arrive at the scene of a crime after it occurs. But rather than send cops to yesterday's
crime, a new trend in law enforcement is using computers to predict where tomorrow's crimes will be -
and then try to head them off.
The software uses past statistics to project where crime is moving. Police in Los Angeles say it's worked
well in predicting property crimes there. Now Seattle is about to expand it for use in predicting gun
violence.
It all started as a research project. Jeff Brantingham, an anthropologist at UCLA, wanted to see if
computers could model future crime the same way they model earthquake aftershocks. Turns out they
can.
"It predicts sort of twice as much crime as any other existing system, even going head-to-head with a
.crime analyst," Brantingham says.
Checking The Boxes
Older systems, like the famous CompStat in New York, show where crime has been. This system looks
forward.
"The model will actually predict other locations, that effectively say, even though there was a crime
somewhere else in your environment, the risk is still greatest in this location today for the next 1 0 hours 01
the next 12 hours," Brantingham explains.
http://sfmail04.sfgov.org/mail/gsuhr.nsf/(%24Inbox)/BC09B9F96736201 041211 CF7343CB !... 9/23113
Seattle police officer Philip Monzon patrols
an area where the department's predictive
policing software has indicated car thefts
are likely to occur.
Marlin KasteiNPR
commit a future crime, just where it is likely to happen.
Page 2 of 4
In Seattle, police Sgt. Christi Robbin zooms in on a map of the city. Earlier this year, Seattle started using
PredPol to predict property crimes. It's now the first place to try predicting gun violence with the software.
"These red boxes [on the map] are predictions of where the next crimes are likely to occur," Robbin
explains.
At the start of every shi.ft, patrol cops are assigned to those red boxes. "So we're asking that they spent
the time in that 500-by-500-square-foot box, doing whatever proactive work they can to prevent that
crime," Robbin says.
On a recent shift, officer Philip Monzon pulls up inside his box; today, it's a city block near the Seattle
waterfront.
"[The police] want visibility, they want contacts with businesses as are appropriate, and anyone who's
wandering through the area," Monzon explains.
This area has parking lots, and PredPol's forecast includes car thefts. As Monzon passes a green Honda,
he pauses. The guy inside seems to be ducking under the dashboard.
"[I] wanna make sure to see if he's got the key or if he's gonna pull out anytime soon," Monzon says.
http://sfmail04.sfgov.org/mail/gsuhr.nsf/(%24Inbox)/BC09B9F9673620 I 041211 CF7343CB 1 ... 9/23/13
Page 3 of 4
The car starts- the guy probably does have the key. But why didn't Monzon challenge him, just in case?
"I don't really have enough- I'm not just going to single out one guy in a Honda," he explains.
Computer Models And 'Reasonable Suspicion'
And this is where this gets tricky. The courts say police need "reasonable suspicion" in order to stop
somebody. That suspicion can come from a lot of things- even someone's "furtive movements," as
police like to say.
All Tech Considered
Police May Know Exactly Where You
Were Last Tuesday
Around the Nation
At LAPD, Predicting Crimes Before
They Happen
But can it come from the fact that someone is occupying an imaginary red box drawn by a computer?
"Ah- no. No. I don't know. I wouldn't make a stop solely on that," Monzon says.
That's probably the right answer, says Andrew Guthrie Ferguson, a law professor at the University of the
District of Columbia who has taken a special interest in the constitutional implications of PredPol. He says
the departments using it have told police not to use it as a basis for stops. But he also wonders how long
that can last.
"The idea that you wouldn't use something that is actually part of the officer's suspicion and not put that
in- [that] may come to a head when that officer is testifying," Ferguson says. Either that officer will have
to omit the fact that he or she was prompted by PredPol, he says, or that officer will admit it on the
stand. "Then the issue will be raised for the court to address."
And it may be that PredPol is a constitutional basis for stopping someone. Some might consider it more
objective than an individual police officer's judgment- less prone to racism or other kinds of profiling, for
example.
Ferguson says that argument may have merit, but that police and society still need to be careful.
"I think most people are gonna defer to the black box," he says. "Which means we need to focus on
http://sfmail04.sfgov.org/mail/gsuhr.nsf/(%24Inbox)/BC09B9F96736201 041211 CF7343CB I... 9123113
Page 4 of4
what's going into that black box, how accurate it is, and what transparency and accountability measures
we have [for] it."
In other words, even though computers aren't biased, the statistics feeding it might be. And if police are
going to follow an algorithm, we should at least be willing to check the math.
DONNIE FOWLER
331 Soquel Ave. Ste. 100
Santa Cruz, CA 95062
PREDPOL.COM
The Predictive Policing Company
http://sfmai104.sfgov.org/mail/gsuhr.nsf/(%24Inbox)/BC09B9F9673620 I 041211 CF7343CB I... 9/23/13
Page I of2
From: Donnie Fowler <donnie@predpol.com>
To: greg.suhr@sfgov.org, Susan.Giffln@sfgov.org, suzyloftus@hotmail.com
Date: Monday, August 12, 2013 08:34PM
Subject:
Georgia City Gets Two Arrests on First Day Using PredPol
Chief Suhr, Susan, and Suzy -
Looking forward to seeing you Tuesday in Berkeley. I thought you would like to see a report from
our first Southern city -- Norcross, Georgia -- to put the predictive technology in the hands of
their patrol officers. Following trainings last Thursday afternoon and Friday morning, this was the
result, as emailed to me by the deputy chief of police.
We look forward to completing our deployment here in San Francisco.
All the best,
Donnie Fowler
4 ell
predool.com
---------- Forwarded message.----------
From: Bill Grogan <bgrogan@norcrosspd.com>
Date: Mon, Aug 12, 2013 at 5,:46 AM
Subject: press release
To: Donnie Fowler <donnie@predpol.com>
Donnie - we want to release something to the media about our PredPol integration. Also, want
to tell you about 2 successes our day shift had Friday. Yes, the same day they received training
2 officers were in a box and caught 2 burglars in a house. Solved several cases with it. Another
officer caught a wanted guy in a box - officer was there spending time because of the box -
arrestee was wanted for probation violation on burglary charges out of Illinois.
So, give me a call please.
Bill
Lieutenant Bill Grogan
Norcross Police Department
Criminal Investigations Division I Crime Suppression Team
bqroqan@norcrosspd.com
770.448.2111 (dispatch)

._(cell)
770.248.1819 (fax)
http ://sfma i I 04 .sfgov .org/ma i 1/ gsuhr.nsf/(%24 In box )/1 AS 9DCC6B3 8A3 3 63 EE8440 B 3 D2 99... 9/23/13
Page 2 of2
facebook
(LESIFOUO) LAW ENFORCEMENT SENSITIVE· FOR OFFICIAL USE ONLY WARNING:
The information contained in this email is considered confidential and sensitive in nature as well as sensitive but unclassified andlor legally
privileged information. It is not to be released to the media, the general public, or to non-law enforcement personnel who do not have
a "need-to-know". This information is not to be posted on the Internet, or disseminated through unsecured channels, and is intended for law
enforcement personnel only. lt is solely for the use of the intended recipient(s). Unauthorized interception, review, use or disclosure is
prohibited and may violate applicable Jaws including the Electronic Communications Privacy Act. lf you are not the intended recipient,
please contact the sender and destroy all copies of the communication.
DONNIE FOWLER Director of Business Development
4104 24th Street, #445
San Francisco, CA 94114
331 Soquel Ave. Ste. 100
Santa Cruz, CA 95062
PREDPOL.COM
The Predictive Policing Company
Attachments:
White Paper Predicting Gun Violence (2013
July).pdf
PredPol Info Sheet How it Works (2013
July).pdf
http://sfmail04.sfgov.org/mail/gsuhr.nsf/(%24 lnbox)/1 A59DCC6B38A3363 EE8440B3D299. .. 9/23/13
' ' c ~ · · · d · ...••c.c················
'
an, a talL
of the Los
P ent. gazes out
ci ;." of black-and-white
·i shakes his head with a
re of disbelief and disdain.
\X!t:'rc JMrkcd on :t quier residential in :\onh ·
I loliv'wood. all :-.inek-ftmih· lwmo, '>h<Hk trct:;. and
tiLh- !.nvm. It\ the sunny day.
no one ;lnd nothing in siglH tlut looks
In f:1ct. isn't ;myonc in sight ;Hall. 'T
don't know wlw we're here," Alharr;m mutttr'i.
Alb:1rran been disrnrchcd w thi.\ unlikely spor
bt'(iltlSt' of thl..' work or ;t young. :lSSiSLHH of
m;.uhcnutics some 3SO ;IW<t)' ;H S;1nt;t Clara
Uniwrsity. George Mohler is a p.tlt:, lnshful :Hl-ye.tr-.
old who happt?ll\ to be hdpin!!. w nne ot
rhc most t:tlked-;Jhouc innov:nions in tllodnn Amerk·:tn
crime Along with scvcral othcr at th.:
Lns he is dt·vdoping
one of most promising experinwtH'> iu an
tleld known as "prnliniw The id(,.-,,: r\ldmush
no one (;l!l know for sure wh<:n an individual mig.ht
(ommit a crime, it co forcc.b( of
and when art" lil.;dy tn be burgled or CH:i
stnlt'!l hv :malwing truckloads or past crime repon.'l :tnd
t)[hcr th;t:l wit.h .,of1hiqkatcd Lumputd ·1 he
;t\t>orirhm :md his colk·ague;, haw
j., work or lwndrcd\ or polk<.'
:tcro">s Los A;·1gdc.-: and in Santa Cruz----:111d yit:lding
impre,,sivc
"\'1/c r:tnk ;ucas :Kcording to rhk," say:-. \lohkr.
identiA,ing whirh are likdie"t to a
em help !;olin• out tn Jcp!oy nftlccr5 w
prcVI.'!H the from h:lppening in tlw plat..'1.'.
But whv the .dgorirbm :1 particuLlr <H<:,l----like
thi:; tlui,er i\ort.h Hollywood block ri•d{.\', i\·fnhk·r
;Kknnwledgl..''i, 11m a!wavs "intuitivdv ck,lr.''
. .
lii'
\lohlt•r isn't cx;tcdv in do'>c touch with rhc nw:m
He .:;pt'nds moq of in a spar:-.ely of!kt•
in the of O'Connor I fall on the Mission
( ;H the t:nd of a -;u!Ht"rr;Hlt.':H\ corrido.r hcdcd-.nl
wirh pos{cr'l upcoming nuth confcn.·nct's
and job opi.'nings for L'omputer He's on dlt'
skinnv ,jde of thin. with :111 ovt'rsio: smile dtat
give:-. i1i111 an :dnm-;t :d:uminglr cheerful look. \Vith
!1i.' n.:ct.mgubr J;t'•'l.:-chic (
sneahts, plaid ;lnd tlop ofhlack h:tir, !w cnold he
<l San Francisco de_,igrwr or an indie rockn-
which he has :Krually. In his spare timl". lw p!;Jycd
b:tss in ;1 L-mrpk· The mtt'lic h:1." gone on hiarm
sino.: he ;llH.l Courtney Elkin Mohler, :111
proK.>ssm in SCU's Dcpartml·nt ofThl';Hre ;md L\UKt',
became pan:nn: in 2011. But Ccorgc still ,
m:magl's to on .\btes t{)r somt" rime :I'l p<tn ol
;lll adu!r hocl(t'V
i\·lnhlcl' gor .his • in
(and hb slajHhot tr.tinir;g) in his lndian.t, and
wt:nt on to rlw lJniver->hv of Califot ni:t, Santa B,ub;ua,
whL:rc he nurilCm:nic:t! moJl'!ing of
and tluid-.:. A!'ter gradtwring, IH.' gm :t job ofiCr In that
tldd bur found more intrigued wich a stranger
{)J1C. Two UCLA otmhropo!ogist Jd'frcy
Buntingham and JJ\athcmatici:lrl Andrea Hcn<Jni,
\\Ttl' working with rhc LAP!) to Jevelop algorithms to
pn:dict crime. They s;IW ;\.-lohkr's and w;tnr<.:'d
him t)fl bo:ud; wrm out ofrhe lllathematic;ll
modds 1\lohlcr h,Hl hem working with thar describe
pattern in w.:re to those
the UCLA were us to pn:dict hurgbric».
"I rc.1d their p;lpcrs, and it madt.· ;I lor of scme,"
"! thoughr what they wen· doing
cool." I k rook rhc job.
The tcnn parhcr\'d vears' worth of data from the
L\PD on rhc,
7
rimc ;Inti phtLl' wht.>rt' home ;Jnd GH
buF•Iaries and auto rhd{,.;; had taken plact'. (They
on thost: crimes bcc;mse tht'll' art;> lnts of
tlwm. providing ;l rich (!au ()nc of t!u:ir e:1rly
imighc' w;l'i dnt Lrinw Il.'nds ro bl"gcr crime: l! :l house
broken into, rhc prob:tbi\iry of neighboring
hrokcn inw soon ::lf{er like
;md c.u rhdf.,. ;lrt-• not planned in :llkmn· hut
arc ;-1pponuni,tic ;\ h;ld gu;· an unlocked window
and dmb in. ''Bmgbrs typic1lly don't !r;wd fir. They
ccnd w commit in their own
,\{oh!er. haw ,1 lnr of int{lf!n;ltion: They
know when their ncil!hhors :He ;H work and which
hnttsc<. <ll'l' w gc(. inw. And wlwn the;: sucLCL'tL tht'y
uo bad, .lg;Jin. You sec ir in dw dau." :-..-tapping tho)t'
f
';.lll''l'll\ c'-tn oive I)Oiict an cdt!,r in nut when:· w
< " ' ' 0 "' "' "'
d.:p\oy i.'Xtra ._·,m and cop,, to cud1 bad guy:.---,: or, bcr_r'-"r
wr, ki.'ep them from opening rhar unlocked wmdow Ill
flr.:;t p!a..:e.
)!1 ways, rlw llOtion of prtdicriny_ Wht'!\' crimes
will ktppl'll ha ... nl on wht:l\' rhcy'n: h;lppt·wx! in the
p<lSt is <1bviom. 'I 'hat om· l'\'Cilt the likelihood
o( simi Jar l'\"t'JllS Ol"t't!frillt.; llt:trh\' in <!lld rime is
wdl c.\tabli.,hn! in or her JieJd,. o/ rcs;.·;tr.._·h, In f:1cL .nHt
can it c'.-cr;·wherl' in orJin:try !ik A punch rhrown
in .1 b:lr rhc ..:h;ull·<:s of mnn· puncht.), On.._•
kb\ 1l'atb w ano1her. ;\n:dy\1" ,1ml :lcHlcmics U'>l' rhc
principle more mcrhodicJI!:· ro predict where b.uuna
might hl' or w!kJL' coqH;LHl' tf...·t:w!t'> will
Unt· uf \lohkr\ main comributiom to a new
for pn:dictin· fn lind .tnd :Hbpt :Ul
algoridm1 by to hdp pn:diL"t
wherL' will srrikt' after ,In e,u £hqu,th:.
If thl'rc han: IKt!l .1 lot of on a p.micul.Jr
S!t\'t•r ll.1r the Lbt SO \\'t:t·b., thnt• will prnh:tbly he
dK· I(JI!owing week. ( know th·.u, of cour..;;.:, But
che ide;t en make rhos(' ns.sumplion., ;md
mot\:' accurah' .tnd to turn up p.mnns dut
ITad[[y <lpp<lfl'll!.
Corponrinns lu\'t' long simil:H prl'di(ti\\.'
to ;unkip;He (on.r,umcr demand. tlnding dut
d:ua em yidd tllll'.Xp<..'"cteJ
r\ L1nwus e;.::!Tllple CO ffom \X'al-.\t!n\ an:lly,\is
of wh:H ib ctb!OillCfS ill CO;!StaJ <lrt.JS Stock up Oil
bef(Jrt' hurric:JIW."o. Tht' Hst includes duct t:1pc .md
hun led warer. naturally: but J.bo ,l sut srrawhtrry
Pop-T1ns.
Ana!ping crime Jau can similarly
counrerintuitin· con(!usions. :>.·!o..;r people think f.Ood
lighting makl'\ an :1r.:-a i0r ill.'lf<li\C('. hut
h.tw lin1nd th,n it ;Kndlv intTL'.lSt'S rhc ch;llKt') of
bcin,t.; vicrimi?.ni. It rh;u mugr.ers w,lnt to ht: ·.1bl"'
to sec thdr porcntbl urgct.\ ck:u·ly.
!11 rhe IL\tnl\ wl)rk, dJOugh still in tht·
tlwnJL·ti;.al.'>t,lgt', pmrnp1cd an ;nride in the l.oJ ,i.;;/!,<'lt'.'
finu•. Th.l! CHigiH dh· ·.trrc'ntion of /;lth .t
L'fime .l!l;Jint wirh the s,ul(,} Cruz Polil"l' D..:parullc!lL
"I calkd uP :\-loh!er. wlw had iu:-.r t;lkcn his iob .1\ SaJll;t
Clara," S,l\'' Friend, .:U. "! qjd. 'Wft''lll<lk,, this out of
tht· cbv.;r(,llllll :wd pur ir into rhe field. if you're willing.'''
The tt':lm •lgreed.
brought ;\lnhkr in tu hdp s(:IJ d1c idt'a to
his The met rhc marlwm:Hi(ian with
;H1HHIIH ofhl'tnliSt'llH.'Ilt. "( )ur th·rd p;t!r'
one onlccr ;'vlohkr wht·n I \'isit the dcpartmt'!H\
ht.:.Hlqu.mcr\ one spring <.by. "l·le's hringint, co1duroy
ntks ·,mrnhn. )till. rhc br<1ss bl)ug!n in. "\\
1
t''ve
h.td cnr;, like l"\'Cfl'nne else. an: ,,car..:e,
,. .
and we need to usc them .h possible."
'i·.lr' Sant:t Cruz Policl' Chief Kevin \'og_t:'L "I ir
wou/J be worth rhi'> <l try."
So every wnrl<tLty t(Jr d1e past two }'L'<lrs, Fril·nd ILlS
('Onlt' in t:arly w rvpc dh,' rime and gcncndt·d location
nf dH· JWJS! rccerH hurgLHY :Hld <HHo thd( rcporrs into
dlt' dl·p;tnllll'!H \ com \ yst t'.H1. }.-f ohl c r\ a! i thnt
thc>n crunche:-. thmt' report!\ togalwr with t!w la..,t seven
\'l':l(<;, worrh of crime data and spit\ out,} ll1:lp or S:una
Cn!L wirh I 0 oil it. t";Kh r'epresentinl!> .111 area 100
long .H1d son widt: ·-ahmu h.llf a b!nck. Thov.'
·,tre the hor rhat ,llgorirhm likdico;r rn
rh;H d.1y. The nnp,\ ,11\' hamh:d our tn nHin·r\
at the beginning oft'xh They tbt'
hoxe.; when thd· h.l\"l' ri111c in hetwn;:n <Ktin·
;'\\?l''rc very Ptc.:l.\l:'d with t1K lnHh'>," U)'S (:hid'
Vogel. In 2012 burglarin do\\'n ahnut 7 petu.•nt
to 20 ll .'And the prngum Ius Jr;twn
''' ''y.! .Not
.;r_ime
.. but
where It's
(to'
Algorithms
For predicting
quakes and crime
ike the earthquake version of the
algorithm to predict where. an
aftershock will strike. mathematician
I)Porou f\llnilter's "expectation
maximization" algorithm For predictive policing
models the incidence of two sets phenomena.
The first is the "background rate" oF events that
occur randomly-earthquakes or spontaneous
crimes. Different geographical areas have
different background rates: Some areas have
more geologic Fautts, and some neighborhoods
havE' social and economic Factors that subject
them to more crime.
The second is the "branching structure"
of similar events generated by the random.
spontaneous ones. With earthquakes. that's
aftershocks. With burglaries. it's more burglaries.
which oFten beget still more burglaries, spreading
out Uki? a tree's branches. You can further narrow
the list of likely locations of future crimes by
ruling out areas 11/here they are impossible; you
can't have a residential burglary in the middle of
a park. for Instance.
Mohler and his team analyzed several years
oF data on the time. place, and type of property
crimes In Los Angeles to see the patterns of
where and when they occurred and were followed
by others. (Murder and other violent crimes
are far less frequent and so don't offer enough
data For the algorithm to provide a reasonably
accurate forecast. says Mohler.) From there.
they derived a set of mathematical functions
to predict both the occurrence of spontaneous
crimes and the probabilistic distribution of where
the crime "aftershocks'' are likely to occur. The
result: a literal road map to future crimes. VB
international aw:ntion. J'imt Tn:lg:1zine, NPR. l11c Nr'/1!
h1rl> !Jm,'i. .111d new_, news from a:-> fir away Fr.HlLT
and ( ,;,_.'rm:Jny luve rt·pnncd nn ir, ant! u( other
agt'!Kics, .l'> well <I'> d1t' 1 kp:H rment of Defense.
han: gottt'll in to111.:h.
:\:'ii(;\1•' rn·· J;J(•!j,l ''Li
One u!' rlw mo.-a OlH -or-rown 1.:nps w.ls
L.\PD ( S\:an ;\hlinowski, ;In ,uhktically huih
46--year-old with hair n:cnling t'rom a :-.:un-redd<:!H:d brow.
coin the term "pn:dictiH· policing''
in an inllut'IHi;!l papa Ill· co-·amhon:-d in 2008 with
1hcn-·Los Police Chicl'\\li\liam Bratton.
M:tlinowski wnrknl • ye;H'\ cxcouivc
t)Hlccr m Br.ttton, who is a nt':lf-legl..'nd in American bw-
t•nf(m:eml'lll cird..:,, d1c police on watch
(rime plllJJllllcteJ firs( in New York, then in LA 1-k·
;t\w,tys gt:tting invirt>d w give {alb oil the llnurt· t)r
policing. :md part uf .Malinowski\ job was w br.tin:-wnn
with him <lhout what to One of lkmon's k(·y
inno\'.Hiom w;t:; a cumpmcrized CompStal,
whil h col!ccts n:ports on criml's and rll<lps where
[hL')' Wt'l"(' committed. of tht· system an· now
w.cJ by police America. Thinking about ways to
hui!d on :md improve CompSr;n's dat;Hlrivcn <lppro;Kh,
they C.lnlt' up wid1 the idt:;l (;tnd C;ttchy ride) of
policing wrmt' :tbnm the ronccpt for tlw Oxjim/
fonrntt!
\\lith Branon\ toweling rr:puutinn behind it, dh:'
idt::t c:tughr lire, Soon aftl.:'r tilL' publication, t!H:
Nation;ll lmfitutc ufJusticC' org,miznl :1 omll•rcncc un
predini\'C policing, and the ofJus£icc h:tndcd
OUt !llOI(' than $\ million in seed gf\tll{S W ,l or
police dl'p<tflmclH.\ in pur:-.uing dw idea.
Vo11 in us ;lgcncin are now nying out diffacnt
.lppm,ldws, pulling in ;111 kinds of d:H:l. In Arlin,(!ton,
Texas, cops have neat.:d m;tps on:rlaying
with building code vioLuions. They tinllld
rlur" ;1-'> physk·al decay up . .so do Thty're
findings to dt•ploy polkt.• more d'l1ciendy.
In Univl.:'rsily of crimino!ngist.\
;tnd !ot:a! police are ming bmincss-;Hhllytics sohw;ue w
rom pill' crime- tepons ;tnd layer in v<l!'i:Ihks like wc:uhcr,
lighdng (onditiom, and proximity ro c.:on<.Trt v<:nucs,
along with reporting from PDi\-cquippcd heal
ro tlnd cunnccriom. Tht· noticl.:'d, for imhtncc.
dJ:ll colkgcs' wct:k rdiahly spawns a rash or
hurgbrit\'>. And in :1 Crimt' An;tlysis
Unit idcnritl,s locninns wht't't' gun crimt':. h.w-.· bctn
rcpnrtcd. dlt'n b\.'tor.'i in geogra\Jhic det,Ji!s o11 thing>.
like hu.; rotHl:'i ,md proximity tu p:1rks, li{jllot
,llld pnhlic lihr.HiL'.\, Combining rhat with d;tt;l
en:tble< rhem to predict when cm:tin public p:trb :tnd
otlwr an_'<l" "·ill bL'Conw .JrctJas gun viok·ncc.
Tht' L.\PD, tnl.:':mwhilc. Ius conrinucd working with
1 !w t\\1111 dut ith.:ltKk\ .nUCLA .1nd A·lohkr
:n SCl}. lmpn.'-"sed wirh the their a!gorirhm
'iccmcd to bl' gening in San{a Crnz, :\blinow-.ki got
;1ppro\·;d !0 ptll ir into jH\Kticc in Angelc ... '>t<lning
ht<' in 201 L
Angdc:. a trial on a comph:rcly ditrnem
sc1k·. S:mt",l Cnu. a b111ously town of only
'lS,OOO in 2012 it a tnt:tl nf two homicide,,
l.o:-. Angel..:" i.., a spnwling td. more th:w
.) million, where someoJH.' gets killed almost day.
E1ced with rlh· size of rhe city and irs pulice
t'or(e, 1\blinow<>ki been intmducing pn:dicrivc
polic.ing nne divi.\ion :u :1 1ime. The
whid1 parrols a chunk of the .San Fernando
Va!k.)' rhat to snnw 2tH.OOO }H.'ople. \\"J.'\ tlw
-.:econd, beginning rome .\!ohh:r\ algorithm ,·arly in
sprinf 2() 12.
"Awo theft, burgLuy frnm and rc,idctHi;d
bmgbrics ;H\' do\\ n I(} comp.trt•d to rhc :i<tll'll,.'
period bst )'L\lr," Captain .Justin tells me
tilt" l\ot th Holl)·wood division he.Jdqu:IIT(T\, ;t
spr.l\vling nwderni\t building on a husy,
<1\Tnuc. ""rJ,,H\ prnry inrn:dible. Prt:Jicrivc polidng
isn't a p,lll<Kca, but it i-.: \tlfpri,ingly m.di.d."
!"he program is now runninf!, in ;t h:llldfu! of
divisinm. with tri:lls under way in c.Kh of rhc cir/;; 21
dlvisinm. Crimc has dropped [n the divi\iolh \\-hen: the
i;, .tlrc:Jdy Ncidn:J nor
;H S.\Jlt,t Ch1r.1 and UCL\ an ..· ready to s.ty
that all th:lt t rime rl'dtKtion is dul' to rh,· algorirhm.
"Bill evn_nmc '"n we'rl· nutnh('r.<; and
it in rheir :Ire,\." .\ldinmnki
!"ht· idca·of policing h:1" i!s crirics. Civil
lihcna1 ians ;1!\.' con....-crncd ir L"rJtdd in cxrr.l pollee
nn poor ·,md minority I! L·np
ho!din1; ,1 b:tg .tnd <H ,\
O(l :1 .\t!\'(.'1 tlw :dgnrit!l!ll h:1.\ ,\S ,I !ikl·1y \pO!
f(H· burglar in. he may be more likd;· to '>top :1nd fri..,k
t!w lnii<.Tet, out Andr...:·w Cmhrie h:r,l{U\llll, an
law at rhc l o( dw l
ut"<.:ohllnbi;J, in .1 fl'Cent paper. Thl" officer
c;Hch ;I rhief----or might opL'n himself" up w ,l ch.ng,c o(
ra,;i;d profiling.
Thcr..:- Jl\' also ".otnc· U!Hlt'I"VingJy Rtpon-
Ltw-cntOrcc.:mt•nt cxpt'filllL'IH' unllcr w:1y tiJ;H tbl'
rechnlquc.\ to hrlp dt:!ermine whedu:r an
is !ikdy w commir ;t crime. Pt.·nn..,ylvJnia
probarion :md p:trnlc oftlci:1ls :Hl' working- with
a 11f Pennsylv:1nia who h<L'-
:lll <ligorithm to hdp rhc Ill'
\peciHc rt··off-(-ndint, ari:cr ;\nJ thl'
Dcp.trrment of Hon1ebnd Sccmity i<>
with <1 .sy .... t.:m ·or scnmr, t!wt trade> airport p:l:.'\l'ngcr..,'
hcart nth.') and other phy.\iGd indicttor:; w help
dctcrmi1K who ,!Joukl nw li.H ;HI t'Xtrd
::.earch. ,\lnhl,,[ th:H hi., :dwnithm [nok::.
only at gcoe,r;iphit: .let:'.ts, nvt indi\·iduals. ''\\
1
c don'r
pur demogr.1phio in h) the model,'' he no
individu.1l infi)rnl<ltion b._·i11g fl'l! in."
On a more practic:d lcvd. h;Hd-hc.tded succt
;He .,kt>ptkal <thour dw whnk· no! inn.
Hack in Alhu-r:tll's cruiser, we move on w <lllothn box
indit·.w:d by dK algMithm. <l hi1Kl< on busy- Ventura
Bnukv,ml. "'Thi.., box lH:re ... \,1\'s Alharr.m. jabbing .1
;H rht> map, "if doem't whar ... ·rime o; who
ro w.uch om \'('e know rhis ;1 with a
k>l nf" \tu(( .>,tn!..:n out tJ( p;trk..;d ·c.u:;, \V.:· don't
necd pn.:::dictivc policing ro tdl th;iL
''I pcrsouall:' don't rhink it\ V'.'ry helpful."' lw
wumbl:;:s. ",\ln:-.t nfnty guy\ the .\.lint: w;t)'··· I hl
poin1. chougb. is no! f(Jr A!b;lrLlH to .1
uimt·-·it·, I(H· hi:-. ro stOp orw hom happen
''\\r'e dl)n't see this .l way to :trf('.\1 pcopk· but to deter
nimc," _qys .\·luhln. "The L\PD rrnh.,bl:- ;ll\11111).', the
rtl!)\f snphiqicncd departments in thl' l rnircd St:Ho.:s. ··he
£.UCS 011. 'They do a good joh. hw wc'vc _\'tHl ...·,tn
do Lt. :\lharr:\11 mig!H th;n. Hut polil",.
:Hound dw u.nHllry <lfl' ,\ttl'nrinn. \·lohlc1 and
collt<l)!Uc-. LHHl<.."hcd :1 (ornnh'n:i;l! Ycrsion ot. rhcir
ptof,L!Ill, duhh.:d PH·dPoL in 2012, ;uld h<lW \t} Lu made
\:de:; tO '>;:\·cr.l! po!icL' depoH <!nd ridded inquiriv.
r:-orn score.\ more .. '\) [Hnhi!c tn:hnnlnt-:r w \\'ill
the progr;lm-----:llhl:ht· wc!l.h \\h.u kind
of i" w nl1!cers. For pn.-didiv"
pol"!cin!!,. it SC'<.'!ib, thv futuJ\' i-.: lonking g(lnd, ..ffi;
Page I of2
From: Donnie Fowler <donnie@predpol.com>
To: greg.suhr@sfgov.org, Susan.Giffin@sfgov.org, suzyloftus@hotmail.com
Date: Friday, August 16, 2013 10:46AM
Subject:
Georgia City Gets Two Arrests on First Day Using Predictive Policing
Chief Suhr, Susan, and Suzy -
Hello again. I thought you would like to see a report from our first Southern city -- Norcross,
Georgia -- to put the predictive technology in the hands of their patrol officers. Following
trainings last Thursday afternoon and Friday morning, this was the result, as emailed to me by
the deputy chief of police ...
We look forward to completing our deployment here in San Francisco.
All the best,
Donnie Fowler
4L cell
predpol.com
---------- Forwarded message ----------
From: Bill Grogan <bqrogan@norcrosspd.com>
Date: Mon, Aug 12, 20p at 5:46AM
Subject: press release ·
To: Donnie Fowler <donnie@predpol.com>
Donnie - we want to release something to the media about our PredPol integration. Also, want
to tell you about 2 successes our day shift had Friday. Yes, the same day they received training
2 officers were in a box and caught 2 burglars in a house. Solved several cases with it. Another
officer caught a wanted guy in a box - officer was there spending time because of the box -
arrestee was wanted for probation violation on burglary charges out of Illinois.
So, give me a call please.
Bill
Lieutenant Bill Grogan
Norcross Police Department
Criminal Investigations Division I Crime Suppression Team
bqroqan@norcrosspd.com
770.448.2111 (dispatch)
)
http://sfmai 104.sfgov .org/mail/gsuhr.nsf/(%241nbox)/E4 E590E6BFFC6C355CB9170E2A26F... 9/23/13
Page 2 of2
facebook
(LESIFOUO) LAW ENFORCEMENT SENSITIVE- FOR OFFICIAL USE ONLY WARNING:
The information contained in this email is considered confidential and sensitive in nature as well as sensitive but unclassified and/or legally
privileged information. It is not to be released to the media, the general public, or to non-law enforcement personnel who do not have
a "need-to-know". This information is not to be posted on the Internet, or disseminated through unsecured channels, and is intended for law
enforcement personnel only. It is solely for the use of the intended recipient{s). Unauthorized interception, review, use or disclosure is
prohibited and may violate applicable laws including the Electronic Communications Privacy Act. If you are not the intended recipient,
please contact the sender and destroy all copies of the communication.
DONNIE FOWLER
331 Soquel Ave. Ste. 100
Santa Cruz, CA 95062
PREDPOL.COM
The Predictive Policing Company
Attachments:
PredPol Reading & News (2013 August).pdf
PredPol Info Sheet How it Works (2013
July).pdf
http://sfmail04.sfgov .org/mail/gsuhr.nsf/(%24 Inbox)/E4 E590E6BFFC6C355CB9170E2A26F ... 9/23/13
DRAFT [May 8, 2013]
PREDICTING GUN VIOLENCE
Cities Take Lead as First in the Nation to Deploy New Technology to Deter Gun Crime
The Challenge. In light of recent gun violence across the country, including school
shootings, and as police departments across the nation face tighter budgets and scarcer
resources, reducing violence and gun related crime remains a significant challenge.
The Leaders. Atlanta, Detroit, and San Francisco are taking the lead as the first cities
to predict and deter gun violence by deploying a new application of PredPol's general
crime prediction methodology.
Predictive Policing. While no one strategy serves as a silver bullet, predictive policing
gives officers a significantly better idea of when and where to be so that they can deter
crime generally and gun violence in particular, sometimes even catching criminals in the
act.
Precursor Crimes. PredPol's unique gun violence prediction methodology recognizes
that past homicides are not necessarily the best predictors of future gun violence. In
fact, the occurrence of serious violent crimes like weapons violations, assault, and
battery provide as much, or more, information on where and when future homicides are
most likely to occur.
Origins & Results of Predictive Policing. Predictive policing was first developed
and deployed in Los Angeles and Santa Cruz, California. Only six months after launch,
those two cities enjoyed declines ranging from -12% to -25% in burglaries, car thefts, and
thefts from motor vehicles compared to the same period in the previous year.
Field Tested. PredPol's general crime prediction technology has been extensively
evaluated using historical crime data and controlled field trials with multiple law
enforcement agencies over several years. Controlled trials show that PredPol predicts
more than h\•o-times as much crime as a veteran police and crime analysts. To be clear,
though, PredPol is not a replacement for veteran officers and crime analysts.
Modeled Using Real Crime Data. PredPol has also extensively modeled its new gun
violence methodology, predicting a greater number of gun homicides compared with
alternative approaches, including traditional hotspot maps. In a test of public crime data
out of Chicago, PredPol successfully predicted so% of gun homicides by flagging only
10.3% of that city-- 66% more than hotspot mapping would. predict.
Science for Patrol Officers. Developments in mathematical and statistical modeling,
high-performance cloud computing, and GPS-enabled mobile devices now make it
possible to deliver real-time crime forecasts on simple maps to patrol officers.
Not Crime Mapping. Without predictive analytics, police end up chasing yesterday's
crime by relying only on intuition and mapping of past crime data. As IAPD's Captain
Sean Malinowski has noted, "We look at these maps and they're as accurate as we can get
them. But I'm looking at a map from last week and the whole assumption is that next
week is like last week" Traditional mapping tools are calibrated less frequently, rely
more on humans to recognize patterns, and allocate resources based on past crimes
rather than predicted future offenses.
Timing. Although the actual deployment date for Atlanta, Detroit, and San Francisco
differ, these cities have smartly added a new tool in the expanding national efforts to
reduce gun violence.
PREDPOL'"
THE PREDICTIVE POLICING COMPANY.™
PredPol Predicts Gun Violence With Open
Government Data
PredPol accurately predicts where and when crimes are most likely to occur. It is the
only predictive analytic system that has repeatedly demonstrated the ability to predict
more than double the amount of crime in head-to-head field deployments against
dedicated crime analysts with all of the tools of the trade. PredPol can now deliver the
same predictive accuracy for gun violence using unique mathematical methods. A study
of Chicago data shows that PredPol successfully predicts so% of gun homicides by
flagging in real-time only of city locations. Knowing where and when gun
homicides are most likely to occur empowers law enforcement to use their knowledge,
skills and experience to disrupt gun crime before it happens.
mile Gllallenge of Gun Miolence - -- ----
In there were 507 homicides and crimes involving handguns in Chicago, Illinois. In
light of recent gun violence across the country including school shootings and as police
departments across the nation face tighter budgets and scarcer resources, reducing violence
and gun related crime from the current high levels observed in cities like Chicago is a
significant challenge.
While no one strategy may serve as a silver bullet, PredPol makes possible the efficient
distribution of limited policing resources. Developments in mathematical and statistical
modeling, high-performance cloud computing, and GPS-enabled mobile devices make it
possible for real-time crime forecasts to be at the disposal of officers in the field. PredPol
technology gives officers the best chance to be in the right place, at the right time, to stop
crime before it occurs.
www.predpol.com PREDICT CRIME IN REAL TIME.™
PREDPOL'"
j
t
.'<(,"'•'"'
Figure 1. Prediction map in 2on: 1% (red boxes), 2% (red and orange boxes), and 3% (red, orange and
yellow boxes) of Chicago flagged, corresponding to 6km\ 12km
1
, and 18km
2
. The percentage of
homicides predicted at thresholds are 10%, 13%, and 20%.
The purpose of this white paper is to show how big data, predictive analytiCs, and hotspot
policing are currently used in practice, and how they are adapted for the purpose of
suppressing gun violence in a city like Chicago. Our methodology allows for several years of
crime data and many different crime types to be systematically combined to yield accurate,
real-time crime predictions. PredPol predictions provide tactically clear recommendations
about where and when to deploy precious police resources. We illustrate the methodology
with a large, open-source data set from the Chicago Police Department.
Success in tl'ie EieiCJ ·
PredPol has been extensively evaluated using historical crime data and controlled field trials
with multiple law enforcement agencies. Analyses demonstrate that PredPol outperforms
existing methods for forecasting crime such as kernel density estimation, which underlies most
crime hot spot mapping programs. Controlled trials also show that PredPol predicts more than
two-times as much crime as a trained crime analyst.
www.predpol.com PREDICT CRIME IN REAL TIME.™
PREDPOL"
<
'
l
<
J
i
J
'
j
'
;
'
~ ~
,
2
, ,;, ,.,\
I
;;
I
>;'''
~
I>';;
I
ooll
I

lh
I
0
·;
~ ~ · ..
,,
-

IIIII
I
I
I
-
I
I
I

·0
0
!! •!O't ·•:'1
m
·•0 c

6 •lOOlt ·lo'-1
"
·>s;,
·•so;,
-n%
·r-H
-25%
"'
"'
serious property crime burglary
Figure 2. Property crime rates fell during the period of PredPol deployment in LAPD's Foothill Division.
Serious property crimes include burglary
1
car theft and burglary/theft from vehicle. Comparisons are
between Foothill Division and six other geographically adjacent Divisions in LAPD.
The PredPol technology offers microscale, real-time geospatial intelligence, displayed
in a way that is tactically clear to supervisors and patrol officers; unlike current models
such as hotspot maps that are often underused because they are ambiguous and
confusing. Over the course of a six-month period of deployment from November 2011
to May 2012 with the Los Angeles Police Department (LAPD), patrol officers nearly
doubled the amount of time on predictive missions from 48 hours per week to 88 hours
per week. Setting clear mission expectations and regular reinforcement of mission
priorities from supervisors were important during this process.
Crime rates can be impacted significantly by the use of PredPol. In the two longest-
standing deployments, with the Santa Cruz and Los Angeles Police Departments,
declines ranging from -12% to -25% during the same period in the previous year were
seen in burglary, car theft and theft from motor vehicle (Figure 2). PredPol serves as a
force-multiplier in allowing officers to use their expert knowledge, skills and experience
in a timely manner in the highest risk locations on the landscape.
BreCiicting Gun Miolenc:e witlil BreCIBol
PredPol is based upon a marked point process methodology that allows for several
years of crime data, and multiple crime types, to be utilized by hotspot maps,
incorporating both fixed risk heterogeneity across the city and temporally dynamic risk.
Chronic hotspots are long term in duration and necessitate problem-oriented policing
strategies to address the root causes of crime. Temporary hotspots, on the other hand,
www.predpol.com PREDICT CRIME IN REAL TIME.™
PREDPOL'"
last on the time scale of days to weeks. Policing strategies must be able to anticipate
emerging trends to disrupt emerging hotspots. Without predictive analytics, officers
end up chasing yesterday's crime. The Los Angeles Police Department updates its
PredPol models for every 8-hour shift and directs patrols accordingly.
In many instances there may be little difference in the situation and intent separating
gun violence and a homicide. The occurrence of serious violent crimes may provide as
much, or more, information on where and when homicides are most likely to occur as
actual homicides themselves. We take the following marked point process approach to
modeling the intensity of homicides. Given marks, M, representing crime types
believed to be precursory to homicide, the intensity of homicide is modeled as:
A.x,y,t=rtx,y+t>tig(x-xi,y-yi,Hi,Mi)
The background rate flX,y represents fixed risk across the city, whereas the kernel
g(x,y,t) determines the time and spatial scales over which near-repeat crime patterns
occur.
. .
€Hicago Gun.€rirne &. Gun ""'orniciCfe
We apply PredPol to an open source data set consisting of 38,740 violent crimes
occurring in Chicago, Illinois in the years 2oog, 2010, and 2011. In total there are 1,331
homicides and the following gun related crimes with "handgun" in the description
field: 17,020 robberies, 6,560 assaults, 8,252 weapons violations, 5,274 batteries, and
303 criminal sexual assaults. The data can be downloaded from the website
· · https://data. c ityofc hicag o. o rg/Pu b I i c-Safety/Crimes-2001-to- prese nt/ijzp- q 8t2".
Table 1. Most significant precursor crimes to PredPol predictions of gun
homicides.
Homicide Robbery Assault Weapons Battery
sth 1St 2nd
Crimes involving guns continue to have an impact on future gun homicides for 30-100
days and risk spreads over as much as 1/2 mile in area. With knowledge of the increase
in crime risk following precursory gun-related crimes, officers are in a better position to
deter more serious gun crimes through directed patrol. In Table 1, we display the crime
types that comprise the highest portion of the estimated risk of homicide and thus play
the largest role in predicting homicide. We note that past homicides are not the
highest predictor of future homicides rather weapons violations, batteries, and then
assaults are the highest predictors in decreasing order. PredPol predictions leverage
www.predpol.com PREDICT CRIME IN REAL TIME.™
PREDPOL'"
these relationships allowing resources to be directed to .where they have the greatest
potential impact.
80
-PredPol
-Random
Predictive AcctJracy
70
-Hotspot Map (3 month)
ii -Hotspot Map (3 week)
.!i
ll60
a

~ 5 { )
:2
E
2 40
"

~ 3 0
~ 20
~
45 67.5 90 112.5 135 157.5
Area of Chicago flagged (km
2
)
Figure 3. Percentage of crime occurring within flagged areas vs. total area of Chicago flagged
with predictions using 2011 historical data. PredPol accurately predicts a greater number of
gun homicides at all levels of deployment compared with standard hotspotting.
PredPol predicts a greater number of gun homicides using its unique prediction
methodology compared with alternative approaches. Figure 3 compares the predictive
accuracy of PredPol against both 3·month and 3-week kernel density hotspot maps. As
police resources increase, more of the city can by flagged for directed patrol and the
number of gun homicides correctly predicted increases. PredPol successfully predicts
so% of gun homicides by flagging in real-time only 10.3% of Chicago.
BreCIBol GuiCies Bolice Bractice
PredPol fits seamlessly into existing police practice and creates strategic and tactical
opportunities that would not exist otherwise. PredPol accurately predicts where and
when gun homicides are most likely to occur. Police can therefore position their
resources to make the best use of officer knowledge, skills and experience to disrupt
gun violence. Moreover, the PredPol methodology creates clear tactical guidelines to
focus on weapons violations, assault and battery as the major drivers of gun homicides.
In a manner similar to 'broken-windows' policing, targeting these precursor crimes can
have a significant impact on gun homicide and ultimately on crime reduction.
www.predpol.com
For more information about this paper or PredPol,
please contact us at: info@predpol.com
PREDICT CRIME IN REAL TIME.™
Q PREDPOL' THE PREDICTIVE POLICING COMPANY."
~
: Rt%eCiie't lt%ime im R.eal mime®
@ ~
Cloud-based, easy-to-use crime prediction software
How PredPol Works
Only three data points from an agency's RMS database:
1) Crime type 2) Crime location 3) Crime time
Predictions are translated onto a map as distinctive red boxes representing an area
of 500 feet by 500 feet. Maps can be broken down into divisions, districts or beats
Predictions are recalibrated for each patrol shift with real-time crime data
Reports are delivered to any Internet-connected device, viewed on an MDT
(Mobile Data Terminal), or printed on paper
How PredPol Benefits Law Enforcement
Twice as accurate as human analysts in Predicting when and where crime will occur
Clear and simple visuals are easy to understand and immediately usable
Complements officer judgment to problem solve while "in/around-the-box", and to
help new officers get up-to-speed more quickly on crime risks in their environment
Identifies precise locations for the most effective use of police resources
PredPol, a force multiplier to disrupt and deter crime
tactical ambiguity tactical clarity
831.331.4550 1 infO@predpol.<om 1 www.predpol.tom
-- -- -
PredPol software uses
a pattern-recognition
algorithm and existing
crime data to make real-
time crime predictions
for safer communities.
Background
PredPol's technology has been
developed over six years by
Internationally recognized PhDs
in mathematics, crimiilology
and anthropology.
The program puts officers on
the scene before crimes occur,
targeting prediction "boxes" as
small as 500 feet by 500 feet.
Successful deployments in
dozens of cities large and small
in the U.S. and abroad.
"''m not going to get more
money I'm not going to
get more cops. I have to be
better at using what I have,
and that's what predictive
policing is about."
-los Angeles Police Chief Charlie Beck
Q PREDPOL' THE PREDICTIVE POLICING COMPANY.''

G ""-"
, Blltellliat im Ileal mime®
" '
i
I
B
•. 0\
: ' .

o
0 • "
Using PredPol, serious property crimes have fallen
by 12% in the Los Angeles Police Department's
Foothill Division.
PredPol is NOT
a replacement for knowledge, skills & experience
profiling of individuals
PredPoiiS
accurate prediction of crime location & time
a focal point for effective local policing
a means for engaging & protecting the public
a way to make best use of your resources
To request a price quote, or for additional information about our custom design
and development services, please contact us at: info@predpol.com or 831.331.4550
831.331.4550 1 fnfo@predpol.com 1 www.predpol.com
User Features
1. https:/1 ... secure login
2. deliver reports on paper, any
Internet connected device or MDT
3. 500' x 500' place-based
prediction boxes
4. crime mapped instantaneously
5. specific to crime type
6. specific to your shifts
7. crime mapping & prediction settings
Applications
Neighborhood policing
Military intelligence and policing
Emergency management and
public health applications
PredPol Requirements
3-10 years of crime data from RMS
database and continuous data updates
Crime type, crime address, and
crime times from RMS database
No new hardware or software required
No incremental training
or maintenance fees
<9 2013 PredPol. Inc. All rights reserved. No part of this publication described herein may be reproduced, stored in a retrieval system, used in a spreadsheet, or transmitted
In any form or by meam eledronic, mechanical, P:hotocopj!ng, recordlng:J or othei'Y!_IS!_ 11er_missLon of PredPol, Inc.
q ' PRE D POL'"
THE PREDICTIVE POLICING COMPANY.'"
Reading & News
predpol.com
PredPol's patent-pending technology generates predictions about which areas and
windows of time are at highest risk for future crimes, including property crime, gun
violence, and traffic incidents. In contrast to "hot spot" analysis that simply maps past
crime data, this technology applies advanced mathematics and computer learning that
has predicted twice as many crimes as those made by experienced crime analysts and
veteran police using existing best practices.
Academic Research:
A screenshot of
PredPol's predictive
boxes that are
provided to patrol
officers on paper or
any mobile device.
Predictions are
specific to iheir shift
and sortable by
crime type.
Please find a link below to ongoing background research that led to the creation of
PredPol, Inc. after successful development and deployment by the Los Angeles and
Santa Cruz police departments. Note that Dr. George Mohler (Santa Clara University)
and Dr. Jeff Brantingham (UCLA) are co-founders of PredPol and remain very active in
its deployments and its ongoing research.
Univ. of California Mathematical & Simulation Modeling of Crime (UC-MaSC)
Dr. Jeff Brantingham: http://paleo.sscnet.ucla.edu/
Publications List: http://paleo.sscnet. ucla .edu/ucmascPubs. htm
I'm not going to get more money. I'm not going to get more cops. I have
to be better at using what I have, and that's what predictive policing is
about ... If this old street cop can change the way that he thinks about
this stuff, then I know that my [officers] can do the same.
· Los Angeles Police Chief Charlie Beck
831.331.4550 info red ol.com www. red ol.c:om
Q PREDPOL'"
PredPol in the News:
THE PREDICTIVE POLICING COMPANY.'"
TIME
~ ~
NBC
( :1 ~ S ®
CBS Evening News- National, March 2012 ("il's not how many people you catch, it's how many
crimes you prevent") I http:llwww.cbsnews.comtvicteolwatch/?ict=7404996n
NPR ·All Things Considered, July 2013 ("Older systems show where crime has been. This
system looks forward.")
h tip: //www. n pr. orq/2 0 13/0 7/26/2058 356 7 41can-softv.la re-that-predicts- crime-pass-co nslituliona 1-muster
The Economist, July 2013 ("Its easier to foresee wrongdoing and spot likely wrongdoers.")
http :1/www .econom is I. com/news/briefinq/2158204 2 -it-q ettinq-e a sie r -foresee-wrong doing-a nd-spot-lik.ely-wro ngdo e rs-d o nt-
even-think-about-it
NBC News- Los Angeles, January 2013 ("a cliff-like drop when predictive policing began")
hllp://www.nbctosangefes.com/news/!ocai/LAPD-Chief-Chartie-Beck-Prediclive-Poticinq-Forecasts-Crime-
185970452.html
Seattle Times, February 2013 ("will allow us to be proactive rather than reactive")
htto://bloqs.seatttetimes.com/today/2013/02/seallfe-police-tUrn-to-computer-software-to-fiqht-crime/
NBC News- Columbia, SC: First City in South to Use Technology to Predict Crime, May
2013 ("going to take us to the next level when it comes to the absence of crime in this city")
http: /lwww. wistv. com/storv/2 227 7055/po!ice-usi nq-tech n o!ogy-to-help-predict-crime
LA, Daily News, Alhambra Cops Test Drive Predictive Policing, May 2013 ("PredPol might
tell us here's an area to focus on today. That may be different from what a police officer knows
based on hot-spot data within the last 30 days.") http://www.dailynews.com/newslci 23243829/cops-test-
drive-preemptive-poHcinq?source=rss&utm source=feedly
BBC - The Age of Big Data, March 2013 ("Crime data represents a treasure trove of potential
information for understanding the nature of crime.")
http://www.bbc.eo.uk/iplayer/episode/b01rt4c7/Horizon 20122013 The Age of Big Data/
MIT's Technology Review, June 2012 ("we are seeing a tipping point")
http://www.technofoqyreview.com/news/428354/!a-cops-embrace-crime-predicling-alqorithml?ref=rss
Associated Press- National, "Sci-fi Policing," June 2012 ("early successes could serve as a
model for cash-strapped agencies")
http://articles.boston com/2012-07-02fbusiness/32493939 1 crime-mapping-patrol-officers-police
Time Magazine - National, "Invention Issue," November 2011 ("getting ahead of the bad guys")
http://www.time.com/lime/maqazine/article/O 9171 2099708-13 OO.html
New York Times, August 2011 ("we have to deploy our resources in a more effective way")
http://www. nytimes.com/2011/08/16/us/16police. html? r=1
news
Q PREDPOL"
THE PREDICTIVE POLICING COMPANY.'"
More Practical & More
Accurate Than Crime Mapping
' I
,, '"'
''
0
' - ,,.,
f
i

'p
-.. ,,
tactical ambiguity
tactical clarity
rear-view mirror heatmap
forward-looking PredPol boxes
Evolution of Policing
This is the next era of policing. , ,
Very soon, we wdl be using a predictive policing model where, by studying real-time cnme patterns, we
can anticipate where a cn·me is likely to occur_
-WilliAm Bull on, • of NYPO mrtd Form.,., Chief oflAPD
LA PO's toolkit now relies heavily on predictive policing.
-Ma)'<X Anto-m-o Vo1hu.ig<lsa ,loiAngef<".s (al!fom;.;,
•,w,w.predpol com
..

' ..
;·fi
,.,
"J
_.,



·"I

'(J

't; ...
•• • "":l
I, 1i [I I

v
.
1];,
,.,·


·L
(fJ
.. .::t
\.:.t·
E
r::.• •• , .•
f l --··
Q PREDPOL' THE PREDICTIVE POLICING
PredPol Predicts Gun Violence With Open
Government Data
PredPol accurately predicts where and when crimes are most likely to occur. It Is the only predictive analytic system that has repeatedly
demonstrated the ability to predict more than double the amount of crime In head-to-head field deployments against dedicated
crime analysts with all of the tools of the trade. PredPol can now deliver the same predictive accuracy for gun violence using unique
mathematical methods. A study of Chicago data shows that PredPol successfully predicts 50% of gun homicides by flagging in real-time
only 10.3% of city locations. Knowing where and when gun homicides are most likely to occur empowers Jaw enforcement to use their
knowledge, skills and experience to disrupt gun crime before it happens.
rl"he Challenge of Gun Miolence
--
In 2012 there were 507 homicides and 12,137 crimes involving handguns in Chicago, Illinois. In light of recent gun viOlence across the
country InclUding school shootings and as police departments across the nation face tighter budgets and scarcer resources, reducing
violence and gun related crime from th_e current high levels observed in cities like Chicago Is a significant challenge:'
While no one strategy may serve as a silver bullet, PredPol makes possible the efficient distribution of limited policing resources.
Developments in mathematical and statistical modeling, high-performance cloud computing, and GPS-enabled mobile devices make it
possible for real-time crime forecasts to be at the disposal of officers in the field. PredPol technology gives officers the best chance to be
in the right place, at the right time, to stop crime before it occurs.
The purpose of this white paper is to show how big data, predictive analytics, and hotspot policing are currently used in practice, and
how they are adapted for the purpose of suppressing gun violence in a city like Chicago. Our methodology allows for several years
of crime data and many different crime types to be systematically combined to yield accurate, real-time crime predictions. PredPol
predictions provide tactically clear recommendations about where and when to deploy precious police resources. We illustrate the
methodology with a large, open-source data set from the Chicago Pollee Department.
>A
.\0

.
l
'
i
!
l
0
r
t ap 0 C··E·\1

.,
""l,
""·'"" \'

, . ...,.,
().,t·..;.•

·,\,,
Figure 1. Prediction map in 2011:1% (red
boxes), 2% (red and orange boxes), and 3%
(red, orange and yellow boxes) of Chicago
flagged, corresponding to 6km
2
, 12km
2
,
and 18km
2
• The percentage of homicides
predicted at these thresholds are 10%,
13%, and 20%.
1 of4
831.331.4550 linfo@predpol.com 1 www.predpol.com PREDICT CRIME IN REAL TIME®
. - - -
Q PREDPOL" THE PREDICTIVE POLICING
J -
' Rr:eCIRol Success in tile liielct -

PredPol has been extensively evaluated using historical crime data and controlled field trials with multiple law enforcement agencies.
Analyses demonstrate that PredPol outperforms existing methods for forecasting crime such as kernel density estimation, which
underlies most crime hot spot mapping programs. Controlled trials also show that PredPol predicts more than two-times as much crime
as a trained crime analyst.
f
'

Ill
'
j
I
!
,
'
;
z
I
{ ,
! '
'
-12%
"'
;
'
,
serious property crime
t

'
z
'

,,,
>cH
,,
o\0 ,
...
. ,.
·>O't
·1;ll
·l<-"1
·!$1>
·)¢"1
'
!
<

i
I
' z
'

-25%
"'
burglary
'
J
I
!
I
Figure 2. Property crime rates fell during
the period of PredPol deployment in
LAP D's Foothill Division. Serious property
crimes include burglary, car theft and
burglary/theft from vehicle. Comparisons
are between Foothill Division and six other
geographically adjacent Divisions in LAPD .
The PredPol technology offers microscale, real-time geospatial intelligence, displayed in a way that is tactically clear to supervisors and
patrol officers; unlike current models such as hotspot maps that are often underused because they are ambiguous and confusing. Over
the course of a six-month period of deployment from November 2011 to May 2012 with the Los Angeles Police Department (LAPD),
patrol officers nearly doubled the amount of time on predictive missions from 48 hours per week to 88 hours per week. Setting clear
mission expectations and regular reinforcement of mission priorities from supervisors were important during this process. Crime rates
can be impacted significantly by the use of PredPol. In the two longest-standing deployments, with the Santa Cruz and Los Angeles
Police Departments, declines ranging from -12% to -25% during the.same period in the previous year were seen In burglary, car theft
and theft from motor vehicle (Figure 2). PredPol serves as a force-multiplier in allowing officers to use their expert knowledge, skills and
experience in a timely manner in the highest risk locations on the landscape.

:_ Rr:ecticting Gun Ntiolence witl:i Rr:eCIRol :""

PredPol is based upon a marked point process methodology that allows for several years of crime data, and multiple crime types, to be
utilized by hotspot maps, incorporating both fixed risk heterogeneity across the city and temporally dynamic risk. Chronic hotspots are
long term in duration and necessitate problem-oriented policing strategies to address the root causes of crime. Temporary hotspots,
on the other hand, last on the time scale of days to weeks. Policlng strategies must be able to anticipate emerging trends to disrupt
emerging hotspots. Without predictive analytics, officers end up chasing yesterday's crime. The los Angeles Police Department updates
its PredPol models for every 8-hour shift and directs patrols accordingly. In many instances there may be little difference in the situation
and intent separating gun violence and a homicide. The occurrence of serious violent crimes may provide as much, or more, information
2 of 4
811,331.45!0 llnf<>@predpol.<oml www.predpol.<om PREDICT CRIME IN REAL TIME<>
-
Q PREDPOL' THE PREDICTIVE POLICING COMPANY."
on where and when homicides are most likely to occur as actual homicides themselves. We take the following marked point process
approach to modeling the intensity of homicides. Given marks, M, representing crime types believed to be precursory to homicide, the
intensity of homicide is modeled as:
;\(x, y, t) = ~ t ( x , y) + L g(x- x
1
, y- y
1
, t- t
1
, M
1
)
t>ti
The background rate JJX,y represents fixed risk across the city, whereas the kernel g(x,y,t) determines the time and spatial scales over
which near.repeat crime patterns occur.
; CHicago Gun €rime 8l Gun l±lomiciCie ·
: -" - - ~ - ~ ~ ~ - ~ ~ - - ~ -
We apply PredPol to an open source data set consisting of 38,740 violent crimes occurring in Chicago, Illinois in the years 2009, 2010, and
2011. In total there are 1,331 homicides and the following gun related crimes with ·'handgun" in the description field: 17,020 robberies,
6,560 assaults, 8,252 weapons violations, 5,274 batteries, and 303 criminal sexual assaults. The data can be downloaded from the
website "h ttps: II data .d tyofch i cago.org/Publ i c-Safety/Crimes-2001-to-p resen tlij zp- q Bt 2 ",
Homicide Robbery Assault Weapons Battery
5th 4th 3rd 1st 2nd
Table 1. Most significant precursor crimes
to PredPol predictions of gun homicides.
Crimes involving guns continue to have an Impact on future gun homicides for 30-100 days and risk spreads over as much as 1/2
mile in area. With knowledge of the increase in crime risk following precursory gun-related crimes, officers are in a better position to
deter more serious.gun crimes through directed patrol. in Table 1, we display the crime types that comprise the highest portion of
the estimated risk of homicide and thus play the largest role in predicting homicide. We note that past homicides are not the highest
predictor of future homicides rather weapons violations, batteries, and then assaults are the highest predictors in decreasing order.
PredPol predictions leverage these relationships allowing resources to be directed to where they have the greatest potential impact.
PredPol predicts a greater number of gun homicides using its unique prediction methodology compared with alternative approaches.
Figure 3 compares the predictive accuracy of PredPol against both 3-month and 3-week kernel density hotspot maps. As police
resources increase, more of the city can by flagged for directed patrol and the number of gun homicides correctly predicted increases.
PredPol successfully predicts 50% of gun homicides by flagging in real-time only 10.3% of Chicago.
3 of 4
831,331.4550 1 in!O@predpol.com 1 www.predpol.com PREDICT CRIME IN REAL TIME®
Q PREDPOL'
__________ __
SO -PredPol
-Random
70
- Hotsp(ll Mop (3 month) i
- Hotspct Map (3 week) !
1'loo
"


2 40
45 67.5 90 112.5
Area or Chicago /lagged (km
2
)
135 157.5
" "
THE PREDICTIVE POLICING
Figure 3. Percentage of crime
occurring within flagged areas vs.
total area of Chicago flagged with
predictions using 2011 historical data.
PredPol accurately predicts a greater
number of gun homicides at all
levels of deployment compared with
standard hotspotting.
----Rrec.fBol Guictes Bolice Rractice ----------- - -
PredPol fits seamlessly into existing police practice and creates strategic and tactical opportunities that would not exist otherwise. PredPol
accurately predicts where and when gun homicides are most likely to occur. Police can therefore position their resources to make the best
use of officer knowledge, skills and experience to disrupt gun violence. Moreover, the PredPol methodology creates clear tactical guidelines
to focus on weapons violations, assault and battery as the major drivers of gun homicides. In a manner similar to 'broken-windows' policing,
targeting these precursor crimes can have a significant impact on gun homicide and ultimately on crime reduction.
For more information about this paper or PredPol, please contact us at: injo@predpof.com or 831.331.4550
4 of4
331.331.4550 llnfo@predpol.tom I www.predpol.tom PREDICT CRIME IN REAL TIME®
© l013 PredPol, Inc. All rigbts reserved'. No part of this publication described herein may be reproduced, stored in a retrieval system, used In a spreadsheet, or transmitted
In any form or by means electronic, mechanical, without the permission of PredPol, Inc.

Sign up to vote on this title
UsefulNot useful