Weekend Learning - Crime detection papers

We know Platir Tech has a crime prediction offering. Some research paper detail out different techniques on crime prediction/detection
Examining Deep Learning Architectures for Crime Classification and Prediction
Key Notes
Inputs for predictive policing

• Who will commit a crime
• Who will be offended
• What type of crime
• In which location
• At what time a new crime will take place

ML Areas

• Predict hotspots
• Predict the possibility for a crime to occur
• Predicting hourly fluctuations in crime rates

Data Patterns

• Spatial and then the temporal patterns
• Temporal and then the spatial patterns
• Temporal and spatial patterns in parallel

Thematic Mapping

• Aggregated to geographic unit areas
• Quick determination of areas with a high incidence of crime
• Visualize geographical extent and duration of crime clusters

ML Features

• 11 crime categories are used as input
• Socioeconomic status and activities
• Data are time-series of data points
• 10 crime types (i.e. “Homicide”, “Robbery”, “Arson”, “Vice”, “Motor Vehicle”, “Narcotics”, “Assault”, “Theft”, “Burglary”,“Other”) 
• Sparsest crime types there is a class imbalance problem
• Data augmentation can come from flipping and rotating the data on their spatial dimension

Paper #2 - Deep Learning for Real-Time Crime Forecasting and Its Ternarization
Key Notes

• Epidemic type aftershock sequence (ETAS) model to crime modeling
• Convolutional neural network (CNN) learn the features for crime forecasting with inputs of historical data, weather, geographical information

ML Features

• Each crime is associated with two times: start and end times
• Crimes in the restricted region
• Temperature, wind speed, and special events, including fog, rain, and thunderstorms
• Trend, period, nearby impact, predict

Britain gave Palantir access to sensitive medical records of Covid-19 patients in £1 deal

Data includes 

• Personal contact details
• Gender
• Race
• Occupation
• Physical and mental health conditions
• Past criminal offenses
• Religious and political affiliation

Happy Learning!!!