Time series analysis and forecasting of crime data
USA has been grappling with crime for decades now and had made significant improvement. However, crime remains to be one of the core societal problems. To build a safer society, we need to take advantage of 21st century’s technology. With current technologies and data availability it is possible to analyze crime patterns and forecast future occurrences of crime. This information is useful for police to increase safety measures and alert the local residents. ‘Predictive policing’ is one such aspect under implementation in few states by the government of USA. This project analyzes and compares the patterns of ‘Chicago’ and ‘Los Angeles’ crime based on history and forecasts future crime rate. These results potentially could help immigrants to choose their area of residence and can help tourists, students and travelers to plan their trips in safer months. In this project, ARIMA, Auto ARIMA, Holts winter and Facebook prophet forecasting models are experimented on Chicago and Los Angeles crime Data. Experimental results show that Holt’s winter and Facebook prophet models give accurate forecasting with Mean Absolute Percentage Error(MAPE) of 9 on one year ahead forecasts.