Prediction of Suspected Criminals Using Weighted Ranking Engine

Abstract

Law enforcement departments are using several criminal databases which contain different parameters related to crimes and criminals. This data is critical for investigation. There is a need to apply data mining techniques on such data to identify interesting patterns and influential parameters. These influential parameters can be used by the law enforcement departments for crime control and investigation. Criminal profiling is used by law enforcement agencies as an investigation tool to predict the suspected offenders and to extract similar patterns which help in prediction of future offences. Different criminal profiling techniques were proposed based on crime nature, geographic locations, physical characteristics, crime scene parameters, modus operandi, victim details, forensic evidence, recency, prolificness and actual location. In Pakistan generally and in the province of Khyber Paktunkhwa particularly, less work has been done in criminal profiling. In this research, a predictive system for criminal profiling has been proposed based on a mathematical model. The mathematical model is constructed from a list of attributes which are critical for criminal profiling. A scoring engine has been designed which assigns scores to criminals on the basis of which they are classified into high-profile (habitual), medium-profile, and low-profile (non-habitual) criminals. This classification will greatly help the law enforcement departments in the process of investigation, prediction and management of criminals. Dataset of prison department of Khyber Pakhtunkhwa has been used in this research. There are thirty attributes available in the dataset from which twenty two have been selected using attribute selection method. Some attributes are derived from the existing attributes and finally a list of attributes has been made for research purposes. The findings of this research shows that five attributes have strong correlation with high act crimes that are: total number of hearings, recovery in high act crime, total number of group members, total number of non-blood visitors and crime frequency. There are three attributes that have strong negative correlation with the high act crimes which are: prison duration, number of dependents and education. On the basis of these attributes a mathematical scoring model is constructed which is used for criminals profiling.