A machine learning approach to predict crime using time and location data

Abstract

Recognizing the patterns of criminal activity of a place is paramount in order to prevent it. Law enforcement agencies can work effectively and respond faster if they have better knowledge about crime patterns in different geological points of a city.The aim of this paper is to use machine learning techniques to classify a criminal incident by type,depending on its occurrence at a given time and location.The experimentation is conducted on a data set containing San Francisco’scrimerecordsfrom2003-2015.For this supervised classification problem, Decision Tree, Gaussian Naive Bayes, k-NN, Logistic Regression, Ada boost, Random Forest classification models were used. As crime categories in the data set are imbalanced, oversampling methods, such as SMOTE and under sampling methods such as Edited NN, Neighborhood Cleaning Rule were used. Solving the imbalanced class problem, the machine learning agent was able to categorize crimes with approximately 81% accuracy.