Indonesian Journal of Electrical Engineering and Computer Science Vol. 22, No. 2, May 2021, pp. 1052~1060 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v22.i2.pp1052-1060  1052 Journal homepage: http://ijeecs.iaescore.com Data mining technique to analyse and predict crime using crime categories and arrest records Most. Rokeya Khatun 1 , Safial Islam Ayon 2 , Md Rahat Hossain 3 , Md. Jaber Alam 4 1,2 Department of CSE, Green University of Bangladesh, Dhaka, Bangladesh 3 Department of ICT, School of Engineering and Technology, Central Queensland University, Australia 4 Faculty of Engineering, Multimedia University, Malaysia Article Info ABSTRACT Article history: Received Oct 12, 2020 Revised Mar 4, 2021 Accepted Mar 21, 2021 Generally, crimes influence organisations as it starts occurring frequently in society. Because of having many dimensions of crime data, it is difficult to mine the available information using off the shelf or statistical data analysis tools. Improving this process will aid the police as well as crime protection agencies to solve the crime rate in a faster period. Also, criminals can often be identified based on crime data. Data mining includes strategies for the convergence of machine learning and database frameworks. Using this concept, we can extract previously unknown useful information and their patterns of occurrence from unstructured data. The sole purpose of this paper is to give an idea of how data mining can be utilised by crime investigation agencies to discover relevant precautionary measures from prediction rates. Data sets are analysed by some supervised classification algorithms, namely decision tree, K-nearest neighbours (KNN), and random forest algorithms. Crime forecasting is done for frequently occurring crimes like robbery, assault, and theft. Specifically, the results indicate the superiority of the random forest algorithm in test accuracy. Keywords: Arrest attribute Crime type Crimes Decision tree K-nearest neighbours Random forest This is an open access article under the CC BY-SA license. Corresponding Author: Md Rahat Hossain College of Information & Communication Technology School of Engineering & Technology | Tertiary Education Division CQUniversity Australia, Building 30/1.12, Bruce Highway North Rockhampton, Queensland, 4701 Email: m.hossain@cqu.edu.au 1. INTRODUCTION Day by day, the crime rate is rising considerably. Crime prediction is not an easy process since it is neither systematic nor random [1]. From crime statistics, some crimes like arson, and burglary, have been decreasing however crimes like murder, gang rape, and sex abuse, have been increasing [2]. Although we cannot predict the crime victims, we can predict the most probable crime locations. Predicting the crime will not completely prevent it from occurring, however, it will offer security to some extent in crime sensitive areas. Usually, people do not take into account the way to be secured from sudden occurrences. Both people who are strangers from outside the area and those already living inside the area should know how and what crime incidents are taking place through that particular area [3]. Varieties of crimes happen in different regions at various times. Life and property of general people can be shattered just because of not having a minimum-security mindset. The mounting crime level has become one of the pressing challenges for society. Police can use crime databases to inspect criminal incidents and associated factors of previous phenomena to implement and