Evaluation of Different Data Mining Methods in Predicting Drug-Related Crimes

Abstract

The prevalence and diversity of drug-related crimes in societies have grown to such an extent that stakeholders involved in combating this issue have been compelled to leverage legal and judicial systems to address it. Consequently, the concept of crime prediction has gradually entered the framework of criminal justice systems. In this regard, criminal justice systems have employed data mining techniques to adopt preventive policies aimed at combating drug-related crimes. The importance of data mining lies in its ability to transform the process of predicting drug-related crimes into a classification problem, wherein various features are utilized to uncover hidden knowledge. Essentially, this knowledge constitutes a classification system designed for the automatic identification of individuals with prior criminal records. This study employs a descriptive-analytical methodology with a quantitative survey approach. The statistical population comprises judicial case files of convicted individuals from the Enforcement of Judgments Office in Shahriar County, collected during the years 2017–2018. A unique dataset was created from this data, dividing the collected features into two classes: individuals with no prior criminal record and those with prior records. The algorithms examined in this study include multilayer perceptron, logistic regression, decision trees, naive Bayes, J48, and self-organizing networks. Data analysis was conducted using the WEKA software. The findings indicate, first, that the naive Bayes classification algorithm outperformed other algorithms in terms of accuracy and efficiency. However, the accuracy levels achieved by other algorithms were also notably high, reflecting the quality of the features. Second, data mining methods can play a significant role in predicting drug-related crimes. Based on the study’s results, it can be concluded that one of the primary methods for crime prediction is uncovering patterns of criminal behavior, a capability that data mining has provided to a considerable extent.