You are on page 1of 29

Sistem Pengambilan Keputusan dalam Perencanaan

Spatial Interaction
Modelling

Dhimas Bayu Anindito, S.T., M.Sc.


Departemen Teknik Arsitektur dan Perencanaan
dhimasbayuanindito@ugm.ac.id

LOCALLY ROOTED,
GLOBALLY RESPECTED
ugm.ac.id
Contents

Basic concept
History of SIM
The utilisation of SIM
SIM equation
Application to do SIM
Apa saja tipe-tipe model? (Minggu 2)
Pemodelan Transportasi

• Bagaimana melihat arus transportasi di sebuah


kota/kabupaten?
• Apa saja atribut dari arus transportasi tersebut?
• Apakah bisa dibuat sebuah model matematis dari arus
transportasi tersebut?
Basic concept: Newton law of gravity
“Every particle attracts every other particle in the
m1
F F
m2 universe with a force that is directly proportional to
the product of their masses and inversely
proportional to the square of the distance between
their centres.”
1
F ∝ m1 F ∝ m2 F ∝
r 𝑟2

𝑚1 . 𝑚2 𝑚1 . 𝑚2
F ∝ F =G
𝑟2 𝑟2
Spatial gravity model

i
Tij

• We divide our spatial system up into N small zones and we refer to these by
indexes such as i and j (e.g. kecamatan within kota, RW within kota)
• Spatial interaction is the flow Tij from one zone to another while if we add
the flows flowing into a zone or out of a zones we get total (aggregate)
activities.
”Everything is related to everything else, but near
things are more related than distant things."

Waldo Tobler’s First Law of Geography


Important figures in the development of SIM

Sir Alan Wilson Sir Michael Batty


Professor of Urban and Regional Systems in Emeritus Professor of Planning in the Centre
the Centre for Advanced Spatial Analysis at for Advanced Spatial Analysis at University
University College London College London
Notation
𝑇11 𝑇12 𝑇13 𝑇14 𝑇15 …….. 𝑇1𝑛
𝑇21 𝑇22 𝑇23 𝑇24 𝑇25 𝑇2𝑛
𝑇31 𝑇32 𝑇33 𝑇34 𝑇35 𝑇3𝑛
𝑇𝑖𝑗 = ෍ 𝑇𝑖𝑗 = 𝑂𝑖
𝑗

𝑇𝑖𝑗

𝑇𝑛1 𝑇𝑛2 𝑇𝑛3 𝑇𝑛4 𝑇𝑛5 …….. 𝑇𝑛𝑛

෍ 𝑇𝑖𝑗 = 𝐷𝑗
𝑖
Mass/attractor

Flow

𝑃𝑖 𝑃𝑗 𝑃𝑖 𝑃𝑗
𝑇𝑖𝑗 ∝ 2 =𝐾 2
𝑑𝑖𝑗 𝑑𝑖𝑗 Parameter

Deterrence

Scaling constant
Generalisation of model
Based on von Thunen model, the deterrence can be substituted by travel
cost (cij). Population (P) can be substituted by any means indicating the
size or activities in Origin (O) & Destination (D), therefore:

𝑂𝑖 𝐷𝑗 𝑂𝑖 𝐷𝑗 −𝛽
𝑇𝑖𝑗 ∝ 𝛽
=𝐾 𝛽
= 𝐾𝑂𝑖 𝐷𝑗 𝑐𝑖𝑗
𝑐𝑖𝑗 𝑐𝑖𝑗

𝜷 is the distance decay parameter


Model calibration
The parameters (K and 𝜷) need to be estimated hence
predicted trips 𝑇𝑖𝑗 are as close as possible to the observed
trips 𝑇𝑖𝑗𝑜𝑏𝑠  model calibration

How? Logarithmic function!


−𝛽
𝑇𝑖𝑗 = 𝐾𝑂𝑖 𝐷𝑗 𝑐𝑖𝑗 → ???
−𝛽
𝑇𝑖𝑗 = 𝐾𝑂𝑖 𝐷𝑗 𝑐𝑖𝑗
𝑇𝑖𝑗 −𝛽 𝑇𝑖𝑗 −𝛽
→ = 𝐾𝑐𝑖𝑗 → log = log(𝐾𝑐𝑖𝑗 )
𝑂𝑖 𝐷𝑗 𝑂 𝑖 𝐷𝑗
𝑇𝑖𝑗 −𝛽
→ log = log 𝐾 + log 𝑐𝑖𝑗
𝑂𝑖 𝐷𝑗
𝑇𝑖𝑗
Linear function: y = ax + b → log = log 𝐾 − 𝛽 log 𝑐𝑖𝑗
𝑂𝑖 𝐷𝑗
Minimum parameters: minimising the sum of squares
(squared deviations) between the predicted and
observed trips

min Φ = min ෍(𝑇𝑖𝑗 − 𝑇𝑖𝑗𝑜𝑏𝑠 )2


𝑖𝑗
But the application of SIM may also give O-D the power function

𝑔 ℎ −𝛽
𝑇𝑖𝑗 = 𝐾𝑂𝑖 𝐷𝑗 𝐷𝑗 𝑐𝑖𝑗
The family of SIM

• Unconstrained
• Singly constrained – Origin
• Singly constrained – Destination
• Doubly constrained
𝑔 ℎ −𝛽
𝑇𝑖𝑗 = 𝐴𝑖 𝑂𝑖 𝐷𝑗 𝐷𝑗 𝑐𝑖𝑗
The utilisation of SIM

• Retail location model


• Modelling population
flows
• Estimating flows of
migration
• Scenario planning

Source: Rae 2009


Modelling population flows
Estimating the flows of migration (Dennett & Wilson 2013)
Estimating the flows of migration (Dennett & Wilson 2013)
Scenario planning

Source: Anindito, 2019

Simulating the mobility flow if three


London boroughs’ median wage are
increased by the development of
creative economy sectors
Then what about the visualisation?
Application to do SIM

• Through R (Example: Dennett, 2018)


• Through Python
Recap

• Spatial interaction model is an application of a famous Physics concept –


other concepts can be applied as well?
• SIM basically measures the interacting flows from existing data; models
and calibrates itself; before it predicts by using other data
• Visualisation matters! What if we want to simulate the transportation
flows in Java island? – A really big dataset
• Contexts can be incorporated into the model
• Creative sources of data can also provide better inputs – a better
decision-making system?
Diskusi lebih lanjut: dhimasbayuanindito@ugm.ac.id

This presentation has been designed using resources from Flaticon.com

You might also like