Criminal Identification, Crime Pattern Detection, and Prediction in India using Data Mining

Crimes in India are increasing at an alarming rate, and criminals are opting for queer activities to commit them. Newspapers, blogs, etc. are day to day filled with various crime incidents. So an approach is proposed for the design and implementation of three different applications: Criminal Identification, Crime Pattern Detection, and Prediction for Indian cities, using data mining. The approach is divided into eight modules� data extraction, data preprocessing, classification and accuracy verification, clustering methods comparison, clustering, map representation, WEKA implementation, and correlation and regression using Rtool. Criminal identification and prediction is done using KNN classification. Clustering comparison is done for three methods: k-means, k-medoid, and agglomerative hierarchical clustering, to find the best method. This proves k-means to be the best method for crime data clustering. Crime detection is done by using k-means clustering, which iteratively generates crime clusters that are based on similar crime attributes. Plotting of crime hotspots using open street map provides visualization to crime hotspots. Verification of k-means results is done using WEKA. WEKA verifies an accuracy of 96% and 97% in the formation of two crime clusters using selected crime attributes. WEKA is also used for other data mining operations such as association rule mining, decision tree, naive bayes prediction, etc. Regression and correlation steps will help to find relations among different attributes. This is done using R Studio. R and WEKA thus helps in prediction. The approach is useful in helping the investigating agencies in crime detection and criminals’ identification, and for predicting crime trends.