DETECTING PICKPOCKETS

FROM LARGE SCALE

TRANSIT RECORDS

Under the Guidance of Mrs.L.PRIYANKA

Submitted by:-
B. TIRU RANGA NAYAKA(20W91A0519)
B.SAI RAKSHITHA(20W91A0523)
B.VAMSHI(20W91A0520)
B.MANISH KUMAR YADAV(20W91A0529)
ABSTRACT:-
Massive data collected by automated fare collection (AFC) systems
provide opportunities for studying traveling behaviors in the urban area.
Existing studies on the AFC data have primarily focused on identifying
passengers’ movement patterns. In this project, however, we creatively
leveraged such data for identifying thieves in the public transit systems. Indeed,
stopping pickpockets in the public transit systems is crucial for improving
passenger satisfaction and public safety. However, it is challenging to find
thieves from regular passengers in practice. For this, we developed a suspect
detection and surveillance system, which can identify pickpocket suspects based
on their daily transit records. Initially, we extract a number of features from each
passenger’s daily activities in the transit systems. Then, in second step we
exploit the strengths of unsupervised outlier detection and supervised
classification models to identify thieves, who exhibit abnormal traveling
behaviors.
INTRODUCTION
In many cities, thefts happen frequently in public transit systems, because
passengers tend to pay less attention to their belongings when they are in a rush
or in a crowded environment. For example ,according to survey in Amsterdam,
Netherlands in the year 2022 there was 7,000 pickpockets caught in subway
system. Many other big cities in the world, such as Barcelona, New York, Rome,
and Paris, are also reported to suffer from the pickpocket problem.
Indeed, it is challenging to detect theft activities committed by cunning
thieves who know how to escape. It is critical to provide a smart surveillance and
tracking tool for the security personnel of the transit systems. With rapid
advances in information technology and data processing capacities, transactional
records collected by automated fare collection (AFC) systems have become
valuable for understanding passengers’ mobility patterns and the urban dynamics.
EXISTING SYSTEM:
The system of existing literature focuses on finding patterns in passenger
activity records. Such knowledge can be useful in a variety of applications, and
plays a vital role in effectively finding and satisfying passenger needs .
We conduct our experiments on real-world datasets containing over
30681 transit records. We split the data into historical training set and evaluation
test set. Specifically, training set covers three months (from April to June, 2022)
and test set comes from the following two weeks (in July 2023). Also, from the
training set, we filter out passengers whose maximum number of daily records is
no more than three. All algorithms and our real-world system were implemented
with Python baselines.
PROPOSED SYSTEM:
In the proposed system, approach is to overcome pickpocket detection problem.
The overall framework of our solution is illustrated in this system. The system first
partitioned the city area into regions with functional categories. Then, the mobility
characteristics of passengers were extracted from transit records dynamically over time.
The core component of the system is made in two steps. The first step being
regular passenger filtering, and the second step being suspect detection. Finally, system
user feedback information, such as newly confirmed thieves, was entered as ground
truth for future model training based.
ADVANTAGES:-
 Effective techniques for Abnormal Traveling Behavior Detection.
 Smart Card Based Transit Record Maintenance.
FIG 1:- Trajectory of a SUSPECT FIG 2:- Trajectory of a regular passenger
System Architecture:
Static data Regular Passenger
Filtering
Map Database
Map
Partition

Functional
POI Regions
Database

No-Suspicion Suspicion
Group Group

Bus Feature
Mobility
Transactions
Extraction
Profiles
Suspect
Detection
Subway
Transactions

Incidence
Reports Feedback User Interaction
Dynamic Data
HARDWARE REQUIREMENTS
 System : i3 or above
 Ram : 4GB or above
 Hard Disk : 40GB or above

SOFTWARE REQUIREMENTS
 Operating System : Windows8 or Above
 Coding Language: Python
 CONCLUSION
In this project, we developed a suspect detection of pickpocket by mining large-
scale transit records. The system assists in identifying pickpocket suspects’ and
enables active surveillance in high-risk areas. Specifically, we first constructed a
feature representation for profiling passengers. Then, we established a novel
two-step framework to distinguish regular passengers from pickpocket suspects.
Finally, we leveraged real-world datasets from multiple sources for model
training and validation, and implemented a prototype system for end users.
Experimental results on real-world data showed the effectiveness of our
proposed approach.