A Machine Learning Approach for Predicting Drug Dependency Risk Among University and College Students in Bangladesh

Drug addiction has become a pressing concern in Bangladesh, especially among the youth. It not only threatens personal well-being but also disrupts relationships, career goals, and overall health. This study set out to develop a predictive model for assessing the risk of drug dependency in Bangladesh by applying machine learning classification algorithms. Drawing insights from journals, research papers, and online reports, several key risk factors related to daily life and socio-economic behavior were identified.To build the dataset, responses were collected from 602 university and college students through a Google Form survey. Of these, 362 participants reported a history of drug addiction, while 240 did not. The questionnaire was designed to capture a wide range of behavioral and socio-economic indicators, with reference to available datasets from the Kaggle community. After data preprocessing in Anaconda Navigator and Jupyter Notebook, a final dataset containing 211 qualitative attributes was prepared.Several machine learning algorithmsincluding Logistic Regression, Decision Tree, RandomForest, Naïve Bayes, and Support Vector Machinewere tested and evaluated using accuracy, precision, recall, and F1-score metrics. Among them, the Multinomial Naïve Bayes model achieved the highest accuracy of 93.02 %, while the Decision Tree classifier reached 84%.The findings highlight that machine learning can be effectively used to predict drug addiction risk by identifying key influencing factors. This approach not only deepens our understanding of addiction but also provides a practical tool to support prevention and intervention efforts in Bangladesh.