The issue of drug-related crimes is a critical challenge for societies worldwide, requiring comprehensive prevention strategies beyond punitive measures. In the Islamic Republic of Iran, traditional criminal justice approaches have proven insufficient in curbing drug-related offenses, necessitating the adoption of advanced predictive and preventive methodologies. This study explores the application of data mining techniques in forecasting and preventing drug-related crimes. By analyzing 1,885 judicial cases from Shahriar’s criminal records, this research assesses the effectiveness of machine learning algorithms in identifying criminal patterns and enabling law enforcement agencies to make data-driven decisions.
Methodology and Data Collection
This study adopts a quantitative research approach, utilizing secondary data analysis. A reverse questionnaire technique was employed to extract structured data from judicial case files. To validate the research instrument, qualitative content validity was applied, while reliability testing indicated a coefficient of 0.80, confirming the consistency of the dataset. A comparative analysis of seven machine learning models, including random forest, naïve Bayes, logistic regression, J48 decision trees, artificial neural networks, fuzzy-neural networks, and multi-stage optimization, was conducted to determine the most effective classification model.
Findings and Discussion
Among the various algorithms tested, the random forest model exhibited the highest classification accuracy, significantly outperforming other predictive models in categorizing drug-related offenses. The study identified gender as the most influential demographic factor in drug-related crimes, with a statistically significant difference observed between male and female offenders. Furthermore, age and education levels played a crucial role in determining the likelihood of criminal involvement, with individuals aged 18-24 and those with a bachelor's degree representing the most prevalent offender demographic.
The analysis of crime duration patterns for different substances revealed varying trends in drug use and criminal behavior. For instance, offenses related to cannabis were more frequently recorded, while crimes involving crack cocaine were among the least prevalent. Additionally, the study found that 51.78% of offenders had no prior record of amphetamine-related crimes, whereas 24.56% had committed cannabis-related offenses within a single day before their arrest. These insights demonstrate the potential of data mining in uncovering hidden patterns and guiding law enforcement in allocating resources effectively.
Theoretical Framework: Proactive Criminal Policy
The study aligns with the theory of proactive criminal policy, which emphasizes preemptive crime prevention rather than reactive punitive measures. The inadequacy of deterrent policies and the ineffectiveness of extreme punitive measures (such as the death penalty) in reducing drug-related crimes highlight the necessity of preemptive interventions. Data mining techniques facilitate such proactive strategies by enabling authorities to identify high-risk individuals and regions, thereby optimizing the deployment of preventive measures and social interventions.
Implications for Law Enforcement and Policy-Making
The integration of data mining in criminal justice systems presents multiple advantages, including:
-Enhanced Predictive Capabilities: The ability to forecast drug-related offenses based on historical data allows for preemptive interventions.
-Cost-Effective Resource Allocation: Targeting high-risk areas and individuals reduces operational costs while maximizing law enforcement efficiency.
-Improved Decision-Making: Data-driven insights enable judicial and law enforcement agencies to formulate evidence-based policies rather than relying on conventional intuition-based methods.
-Support for Rehabilitation Strategies: By identifying offenders at risk of recidivism, rehabilitation programs can be tailored to specific demographic and behavioral patterns.
Conclusion
The findings indicate that machine learning models, particularly random forest algorithms, significantly improve the accuracy of crime forecasting. The study underscores the necessity of data-driven policing as an essential component of crime prevention strategies. Additionally, it advocates for a shift from punitive to restorative justice approaches, where rehabilitative interventions play a key role in reducing drug-related offenses. Future research should explore the integration of real-time data analytics and AI-powered surveillance systems to further enhance crime prediction and prevention capabilities.