Crime Forecasting : A Theoretical Approach

Crime and its increase are detrimental to any society. High crime rates can affect the functioning of a city and its economy. The prediction and future forecasting of crime can help law enforcement officers curb crime in a city. There is a need for a better prediction system which intimates law enforcement regarding hot spots of crime in the city. Previous research shows a lower accuracy in crime prediction and forecasting using machine learning. Therefore, this study includes the application of different Machine Learning models like, Decision Tree, Extra Tree Classifier, K- Nearest Neighbor, AdaBoost, Random Forest, Light Gradient Boosting and Gradient Boosting algorithms and time series analysis by Long Short Term Memory and Convolutional Neural Network, Auto Regressive Integrated Moving Average and Seasonal ARIMA to suit our data.The performance of a fully convolutional neural network showed the most promising results. Moreover, identification of districts which have the most contribution to the total crime is also determined. This study is useful in assisting law enforcement in identification of districts with higher crime and future patterns of crime in the city of Boston.