Crime Prediction Application Using Artificial Intelligence

The identification of patterns amongst crime incidents has been the subject of various studies, and an important challenge in criminology is the prediction of crime occurrences with a high level of accuracy. Currently, the technique of statistical crime hot spotting, alongside empirical knowledge of the law enforcement agencies; is applied. In this paper, we provide a fresh and minimally explored approach at solving this problem, by using an artificial neural network-based model, for the purpose of area-based crime prediction. Our model can predict where crime will be most prevalent, being of obvious benefit to a city and its citizens. In order to create actionable insights, the model that we created is able to predict crime within a small-time frame and a small geographic region. In addition, our experimental results on the Boston Police Department's open data set consisting of crimes across a span of 10 years yielded an accuracy of 81%, which is a marked improvement over recent models proposed in literature.