A Classification Approach for Crime Prediction

Abstract

Crime is a universal social issue that affects a society’s nature of life and economic growth. With ever-increasing crime rates, law enforcement agencies have begun to show interest in data mining approaches to analyze crime patterns in an effort to protect their communities. Existing work in crime prediction is carried out by clustering the attributes into a set of crime categories. This paper is taking the classification approach to predict crime category by building and comparing the performance of two classifiers; Random Forest and Support Vector Machine. The classification model is built using the UCI Crime and Communities dataset that consists of demographic information and other attributes. The results have shown that Random Forest has outperformed the Support Vector Machine in classifying the crimes with an accuracy of 99.9% due to the mixed nature of numerical and categorical features in the datasets.