Crime Analysis and Visualization in Atlanta Based on Spatio-Temporal Kernel Density Estimation (STKDE)

Team 12: Jingjing Ye, Guangyu Min, Ziheng Xiao

The project code contains two parts:

1. Data analysis: including refined data, STKDE output data, prediction data, STKDE code
2. Data Visualization: including front-end visualization result, and user interface

Install the following libraries:

-pandas -numpy -datetime

-scipy -math

-sklearn -torch

-matplotlib -seaborn

To view the visualization result, visit https://kratosst.github.io. For the data presented in the demo, visit the corresponding github page (https://github.com/KratosST/kratosst.github.io) for presented data.

There are several files to process the data. These codes are designed for data merging, refining and scaling.

Crime Map Visualization Interaction:

1.The bottom slider defines the time range from Jan 1, 2020 to Dec 1, 2021. The data presented in the time range from Jan 1, 2021 to Dec 1, 2021 are prediction results. The first half analyzes on previous crime data and aggregates on the density. Draw the slider to select a single day and view the heatmap.

2.The dropdown bar defines the splitted hour range in a single day. Select one period and view its the heatmap data for a specific day chosen in step 1.

3.Move mouse on each neighborhood to view the previous crime analysis.

4.Click on the map to see the crime analysis with prediction on a specific region grid. The grid size is predefined and used in STKDE and prediction.

STKDE:

In file densitySpaceTime, there are 4 files

$ python main.py

main.py will implement stkde and save the density data then use draw class to draw a 4 dimension pic.

Predict:

In file MLP there are two files here density_to_pre.py, mlp_lib.py

$ python density_to_pre.py 

density_to_pre.py will read csv and use the mlp class in mlp_lib.py to do predict