TY - JOUR
T1 - Offender Characterization and Prediction
T2 - A Case Study of The Kingdom of Bahrain
AU - Alasfoor, Ebrahim J.
AU - Alshaikh, Omar
AU - Inuwa-Dutse, Isa
AU - Khan, Saad
AU - Parkinson, Simon
N1 - Publisher Copyright:
© 2013 IEEE.
PY - 2025/1/20
Y1 - 2025/1/20
N2 - Crime analysis using data mining techniques can uncover valuable information to assist law enforcement investigations. This study analyses a crime data set from the Kingdom of Bahrain to demonstrate an integrated framework using exploratory analysis and machine learning for predictive modelling. The study data set, comprising 720 cases, was first explored to reveal patterns, including prevalent theft and assault crimes concentrated in four areas. Data sensitivity measures were implemented by anonymising specific geographic areas, represented by characters A to T. Supervised learning algorithms such as Additive Regression, Random Forest, Naïve Bayes, and Decision Tree are then applied to predict unknown offender characteristics such as age, sex, and occupation. The algorithms achieve strong predictive performance. Additive regression achieved a high correlation coefficient of 0.889 for the prediction of the age of the offender, and Random Forest achieved 83% precision in predicting sex through 10-fold cross-validation. However, challenges arise with unbalanced nationality data. A survey of 22 police officers was conducted, discussing the study results and the feasibility of the predictive modelling approach to enhance investigative efficiency. The responses received were mostly positive, reflecting a strong endorsement of the potential of the proposed method to enhance law enforcement practices. This research highlights a promising methodology that combines exploratory and machine learning techniques to extract actionable insights from crime data. With appropriate caution, it can improve investigative efficiency by making data-driven predictions of the attributes of the offender. Further work should expand data sets and explore advanced algorithms while addressing ethical concerns regarding predictive modelling in criminal justice.
AB - Crime analysis using data mining techniques can uncover valuable information to assist law enforcement investigations. This study analyses a crime data set from the Kingdom of Bahrain to demonstrate an integrated framework using exploratory analysis and machine learning for predictive modelling. The study data set, comprising 720 cases, was first explored to reveal patterns, including prevalent theft and assault crimes concentrated in four areas. Data sensitivity measures were implemented by anonymising specific geographic areas, represented by characters A to T. Supervised learning algorithms such as Additive Regression, Random Forest, Naïve Bayes, and Decision Tree are then applied to predict unknown offender characteristics such as age, sex, and occupation. The algorithms achieve strong predictive performance. Additive regression achieved a high correlation coefficient of 0.889 for the prediction of the age of the offender, and Random Forest achieved 83% precision in predicting sex through 10-fold cross-validation. However, challenges arise with unbalanced nationality data. A survey of 22 police officers was conducted, discussing the study results and the feasibility of the predictive modelling approach to enhance investigative efficiency. The responses received were mostly positive, reflecting a strong endorsement of the potential of the proposed method to enhance law enforcement practices. This research highlights a promising methodology that combines exploratory and machine learning techniques to extract actionable insights from crime data. With appropriate caution, it can improve investigative efficiency by making data-driven predictions of the attributes of the offender. Further work should expand data sets and explore advanced algorithms while addressing ethical concerns regarding predictive modelling in criminal justice.
KW - Crime Analysis
KW - Criminal Investigation
KW - Data Mining
KW - Machine Learning
KW - Predictive modelling
KW - EDA
UR - http://www.scopus.com/inward/record.url?scp=85215857129&partnerID=8YFLogxK
U2 - 10.1109/ACCESS.2025.3531655
DO - 10.1109/ACCESS.2025.3531655
M3 - Article
VL - 13
SP - 29406
EP - 29431
JO - IEEE Access
JF - IEEE Access
SN - 2169-3536
ER -