Report: https://rubenahrens.com/docs/convlstm.pdf

A machine learning model to predict criminal activity in the city Chicago.

Chicago crime data

Light intensity map

Street lights

Alley lights

Weather data

Shapefile

Police Station Locations

Crime severity

Poverty Indicators

• Recreate the model for crime prediction in Chicago
• Add and preprocess only a subset (most important) data
  • create mask to delete predictions outside of Chicago.
  • Use the last 10 years (2011-2021)
• Evaluate the model: MSE & RMSE
  • Baseline historical average
  • LSTM
• Add more data that might improve accuracy
  • street and alley lights
  • demographic data
  • weather
    • convert to (ndays,55,50) sized array
  • severity of crimes
    • column of severity to be used as weight for the average