Serial Commercial Crime and Population Pattern Mining Using K-Means Clustering and K-Star Classi-fication

Abstract

Crime pattern mining has taken a center stage recently that given many governments a push to increase its power to reduce the number of crime and to prevent another case and also apprehend the criminal. In Malaysia, there is yet any research that combines both population and commercial serial crime information to retrieve inquisitive patterns. In data mining, the use of machine learning has played a crucial part not only in obtaining patterns but also understanding the hidden knowledge behind the data. The purpose of this study is to introduce the use of unsupervisedK-Means algorithm to find groups in serialcommercialcrime obtained from Royal Malaysia Police (RMP) and demographic databasesreceived from National Department of Statistics, Malaysia and a set of supervised machine learnings to obtain the best accuracy of crime pattern based on types of crime.This study is based on datasets from Selangor and Kuala Lumpur with 15857 instances ranging from January 2012 until June 2014. The results show that K-means clustering able to repopulate the crime pattern according to its criminal’s demography with Cluster=11 and 17441.0within cluster sum of squared errors whereas in supervised learning, Kstar gained the highest accuracy rate upto 75.34%.