A Machine Learning Model for Predicting Individual Substance Abuse with Associated Risk-Factors

Abstract

Substance abuse is the unrestrained and detrimental use of psychoactive chemical substances, unauthorized drugs, and alcohol that can ultimately lead a human to disastrous consequences. As patients with this behavior display a high value of relapse, the best intervention approach is to prevent it at the very beginning. In this paper, we propose a framework based on machine learning techniques to identify individual vulnerability towards substance abuse by analyzing socio-economic aspects. We have carefully assessed the commonly involved causes to form the questionnaire for collecting data from healthy people and patients suffering from substance abuse. Using Pearson’s chi-squared test of independence, feature importance is measured to eliminate less significant features using backward elimination. Popular machine learning classification algorithms (Random Forest, Decision Tree, Logistic Regression, Support Vector Machine, K-Nearest neighbors, and Gaussian Naive Bayes) are used to build the predictive classifier. To identify the key risk-factors of individual substance abuse, we extract association rules from the significant features and subsequent factors. Experimental results on real data-set support the effectiveness of the proposed framework.