The Usual Suspects: Machine Learning Based Predictive Policing for Criminal Identification

Abstract

Crime is a critical problem faced by almost every state. Particularly, the crime rates in the under developed countries are alarmingly serious. Generally, in under developed countries, the criminal record is maintained manually and linkage between different criminal records is near to impossible. Investigation officers have to manually look for clues and leads to predict potential suspects who may be involved in a recent crime. Mostly the serial offenders follow peculiar methodologies to execute unlawful activities, referred to as the modus operandi. In this paper, we propose a machine learning based predicting policing algorithm exploiting the modus operandi features of a recent crime and the existing prior criminal records. The proposed model predicts and probabilistically shortlist the potential suspects who may be involved in a recent crime. Thereby, meaningful leads are produced to aid the investigation process. In this paper, we take the Punjab province of Pakistan as a case study and apply the proposed model on the real data collected from various police stations. The results show that the identification of potential suspects proved vital for the cases which involves criminals who had previous criminal record.