Early Intervention: A Machine Learning Approach to Classify Drug Addiction

Abstract

Alcohol and drugs are harmful to the body and health. Drug addiction is becoming a menace to the youth of Bangladesh. We will use machine learning to forecast the likelihood of developing a drug addiction according to drugs symptoms. After reading relevant studies, journals, and internet publications and speaking with medical professionals and drug users, we were able to identify a few commonalities in the development of various types of drug addiction class. Then we collect my real data from Divisional Drug Addiction Treatment Centre, Department of Narcotics Control, Rajshahi. almost 21 on those features, such as Age, Gender, Living Situation, Motive of Drug Use, Time Spent Mostly, Failure in Life, Symptoms, Label etc. We collect our data from only addicted people from the agency. 8 classes of drug addicted people data have been collected such as 'Addicted-Heroin', 'Addicted-Alcohol', 'Addicted-Cannabis', 'Addicted-Meth', 'Addicted-Ecstasy', 'Addicted-Prescription Opioids', 'Addicted-Cocaine', 'Addicted-MDMA'. We collected the data, processed it all, and produced a processed dataset. We used machine learning methods on the dataset we had previously processed. Since different prediction and detection systems employ machine learning, artificial intelligence, and deep learning. We employ decision trees, random forests, XG Boost, naïve Bayes, Support Vector Classifier (SVC), and k-nearest neighbor (KNN). Among the six algorithms used in our experiment, decision tree models performed the best in terms of accuracy; the classifier's accuracy was 97.75%. Then create a web application according to the decision tree model based for predict various drug addiction using their symptoms.