sis-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Nadeem Anjum (JIRA)" <>
Subject [jira] [Commented] (SIS-97) [GSoC] Agent Based Modeling based geo-profiling of criminology projects
Date Thu, 02 May 2013 11:06:16 GMT


Nadeem Anjum commented on SIS-97:

Hi Suresh,

The deliverables will be the following: (details of which are in the previous post)
1. A Web Interface with Google Maps Embedded in it
2. Randomizing Agent Moves on the Map
3. A Memory Function
4. Introducing new features including maps, movement of v-agents, movement of police, increased
probability of committing a crime in an unoccupied spot, in a region where previous break-ins
have been successful, in regions where patrolling police are far away, associating a profit
value with each crime spot, associating a fraction denoting the ease of breaking into a house.

Predicting Crime Patterns by:
5. Simulating Distance Decay Theory on the web interface with maps and other new features
mentioned above
6. Simulating Routine Activity Theory on the web interface with maps and other new features
mentioned above

Additions to what is already in the paper:

1. The c-agent moves along roads in the map, rather than along a simple 2D grid. Only legal
movements will be allowed. For e.g movement allowed only in one direction along a one-way
2. At the starting point, the c-agent decides to left or right on the road randomly with equal
probability of 0.5
3. At each road junction, the c-agent decides to choose a road randomly with equal probability.
If there are x roads on an junction, the c-agent chooses a road with a probability 1/x
4. The crime spots can be houses, buildings, parks, shops etc. i.e. any establishment on the
map. Each of these has a v-agent associated with it.
5. Each crime spot has a different profit value, rather  than all having the same profit value
and the probability of breaking into a spot is directly proportional to the profit value.
6. The c-agent recognizes all v-agents falling with a fixed circular area around it in the
map (say 100 metres square), rather than recognizing only adjacent cells
7. The memory function decays with the distance traveled by the c-agent, assuming that the
c-agent moves at a constant speed.
8. Each crime spot has a fraction p associated with it, which denotes the ease of breaking
into the spot. The probability that the c-agent breaks into the spot is directly proportional
to the value of p. This fraction will consider various factors like security cameras, lights,
fences, strong locks, time of the day (for e.g ease of breaking will be higher at night than
9. We shall also simulate movement of v-agents: Each v-agent in its territory decides periodically
with a certain probability whether or not to move out. If it moves out, it moves randomly
similar to the c-agent, with the addition that it periodically decides with another probability
whether or not to go back home. It moves back home via the shortest path. The probability
of a c-agent breaking into an empty household will be kept high, giving the c-agent an option
to break into an unoccupied immediate neighborhood too.
10. We also introduce patrolling by the police. The police is made to move randomly similar
to c-agents and v-agents and c-agents. The probability that the c-agent breaks into a house
will be inversely proportional to the distance of a police from the house. If the c-agent
breaks into a house with a police within a threshold distance, it is caught.
11. The probability of committing crime in a particular region will increase with each successful
crime in a region (say around 100 meters square of a successful crime spot).

> [GSoC] Agent Based Modeling based geo-profiling of criminology projects
> -----------------------------------------------------------------------
>                 Key: SIS-97
>                 URL:
>             Project: Spatial Information Systems
>          Issue Type: New Feature
>            Reporter: Suresh Marru
>              Labels: gsoc2013, mentor
> This idea targeted for GSoC project to extend the geo-profiling computational criminology
projects discussed in [1], [2]. Paper [2] uses a simple 2 dimensional spacial grid with x,
y co-ordinates to move the agents. This project should instead use google maps api to make
the agents move around as we validate various criminology theories like Distance Decay and
Routine Activity [3]. In the future these could be modeling in SIS when rendering is fully
> The project involves engaging with the SIS community for special expertise, understanding
google maps api, understand the basic concepts of agent based modeling using a simple proof
of concept like example [4]. The project deliverables includes a web based interface embedding
google maps and randomizing agent moves and develop a memory function and overlay mocked crime
data and predict crime patterns. 
> In addition to SIS PMC, external backup mentors Suresh Marru, Ramyaa Ramyaa and Arvind
Verma will provide the necessary guidance on the theory of computing  and logic, agent based
modeling and criminology concepts. 
> This projects is suited for students who can interpret computer theory and logic and
efficiently implement algorithms, to prototype with GIS based maps and have enthusiasm to
publish multi-disciplinary conference papers. 
> [1] -
> [2] -
> [3] -
> [4] -

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see:

View raw message