Predicting Drug Addiction Patterns in Urbanizing Bangladesh: A Machine Learning Approach

Abstract

Drug addiction is a significant public health challenge worldwide, with its impact on indi- viduals, families, and communities increasing in fast urbanizing countries. In Bangladesh,

the rapid growth of cities, socio-economic inequities, cultural shifts, and increased drug ac- cessibility have led to complicated addiction patterns that require quick treatment. Factors

such as peer pressure, unemployment, mental health issues, and family history of substance

use play crucial roles in the onset and progression of addiction. Machine learning technolo- gies have emerged as strong tools in forecasting complicated human behaviors, including

addiction risks. This study uses Logistic Regression, Decision Trees, Random Forest, Sup- port Vector Machine, Naive Bayes, AdaBoost, KNN, and MLPClassifier to find significant

factors of drug addiction in Bangladesh and construct a risk prediction model. This model assesses the probability of addiction at individual and community levels, providing data for politicians and healthcare practitioners to plan focused therapies. these are result of ten models. Logistic Regression result is 0.80, Decision Tree result is 0.77, Random Forest result is 0.79, SVM result is 0.80, Naive Bayes result is 0.80, AdaBoost result is 0.81, KNN

result is 0.77, MLPClassifier result is 0.79, Gradient Boosting result is 0.80, CatBoost re- sult is 0.81This research fills a vital information gap by blending demographic, behavioral,

and familial data to offer actionable insights. By adopting a data-driven approach, this study aims to provide a deeper knowledge of the distinct risk profiles of individuals and communities in Bangladesh. The use of predictive modeling methods can offer a more exact way to quantify addiction risks, enabling healthcare practitioners and governments to spend resources more effectively and devise focused therapies. The validated predictive model will evaluate the chance of drug addiction for individuals in Bangladesh, offering a risk score ranging from 0 to 100% and dividing individuals into low-risk and high-risk groups based on important demographic, psychological, and environmental factors. The findings will enable the development of intervention programs and evidence-based policy recommendations aimed at lowering addiction risks. In conclusion, this study serves as a significant step in addressing drug addiction in Bangladesh, producing a healthier and more supportive society.