CRIME PATTERN DETECTION USING DATA MINING TECHNIQUES: CASE OF SHIRE TOWN POLICE OFFICE

Abstract

Shire Town Police Office is not using any technology based system to make analysis of criminals’ activities to understand trends of previous years’ crimes and to identify the prevalent crime patterns occurred. This problem is not impacted only for the Shire Town Police Office but also for the region and the country. This research investigates the potential of data mining tools and techniques in developing models for crime pattern analysis to support the crime detection activities at the Shire Town Police Office-Tigray-Ethiopia. Out of more than 10,000 offenders’ record the researcher used only 9967 offenders’ recordS and 11 attributes data for this research. Utilizing clustering and classification algorithms, specifically K-means for clustering, J48 Decision Tree and NaïveBayes for classification, the research analyzes real offenders' data collected from the Shire Town police office. The results demonstrate that the J48 Decision Tree model, achieving an accuracy of 97.48% with 119 Number of Leaves and 157 Size of the tree, outperforms other models in detecting crime patterns based on the criteria that classifiers evaluated. Based on the findings of the J48 Decision Tree one sample is listed like: if the occurred crime type is at 2007 e.c, offenders who are grouped in the age group of Age2 (20 up to 32 years old), their Educational status is illiterate, and the crime occurred time is AM, then 109 offenders (97.32% of them) are classified as Male. There are 112 records. From which 3 records (2.67% of them) are incorrectly classified. This study highlights the significance of data mining in transforming raw crime data into actionable insights, thereby facilitating more effective decision-making in law enforcement. The findings emphasize the necessity of implementing modern data mining techniques to improve crime management strategies, ultimately contributing to enhanced public safety in the region.