P-P-P

Spatio-Temporal Modeling: Point process prediction for mortals

Silver Medal Winner for STEAM $2500 Prize! 🎉 🎉 🎊

Code for Project

Introduction

Burglaries, earthquakes, and tweets all have a particular characteristic in common. The occurrence of one event increases the probability of subsequent events. Earthquakes can produce aftershocks,tweets can produce subsequent re-tweets, and burglaries follow the same behavior.

Self Exciting Point Processes(SEPP) models are built with this behavior in mind and they have been tested to be 1.5 - 2.2 times more accurate than previous approaches. This is an open source implementation of SEPP technology for police departments.

How to use it?

1.Find crime data of your city. Example

2.Import data to R.

3.Run our code (pdf version)

Our UI:

Extra Visualization(Carto):

Helpful Links

UCLA Statistics work: http://www.stat.ucla.edu/~frederic/papers/crime1.pdf

Our Report: https://github.com/Italosayan/P-P-P/blob/master/Burglary%20Pattern%20Prediction%20Report.pdf

Our R code: http://htmlpreview.github.io/?https://github.com/Italosayan/P-P-P/blob/master/Analysis_Code/final_analysis.html

Slides Presentation: https://github.com/Italosayan/P-P-P/blob/master/Crime%20Pattern%20Prediction%20Presentation.pdf

Web App code : https://github.com/Italosayan/P-P-P/tree/master/MapApp

Download visualization of the San Antonio dataset: https://github.com/Italosayan/P-P-P/blob/master/Graphics/crimedataset.mov

Mohler's explanation: https://vimeo.com/50315082

Visualizations

G Function Distribution San Antonio Data:

U Function Distribution San Antonio Data:

Lambda Function Distribution San Antonio Data:

We choose the points with the highest lambda value as the risky ones.

Contributors:

Italo Sayan: ixs3409@rit.edu

Nathan Raw: nxr9266@rit.edu