A Comparative Machine Learning Study to Predict Drug Addiction in Bangladesh

Abstract

Drug Addiction is one of the growing threats all over the world. According to Dhaka Tribune, more than 7.5 million people are addicted to drugs in Bangladesh. There are a lot of differences between a drug-addicted and a non-addicted person on health condition, social life, personal life, and familial life behaviors. So, steps should be taken to prevent drug addiction with proper curative issues. In this paper, we dig for the influential factors behind drug addiction and possible solutions to reduce the drug addiction rate. The research is held on the people of Dhaka, Bangladesh. Most of the data of drug-addicted people are collected from ‘Drug Rehab’ and for non-addicted person data we have collected from different schools, colleges, and universities in Dhaka, Bangladesh. All are male and the age group of 17 to 45 years. Our primary data set is constructed including only 188 qualitative data. A total of 5 algorithms have been employed including Logistic Regression, Decision Tree, Random Forest, Naive Bayes, Support Vector Machine (SVM) and their results are compared. Among the algorithms Random Forest comes up with the highest accuracy of 97.3484%, XGBoost & Decision Tree Classifier delivers the accuracy of 96.2768% and 94.68%.