A Machine Learning Approach to Predict Volatile Substance Abuse for Drug Risk Analysis

In this work a machine learning approach is proposed for prediction of volatile substance abuse. Machine learning technique used in this work is artificial neural networks (ANN). Two ANN modules are designed, ANN-D to predict whether a person is using VSA or not and ANN-C to predict the time of use. Input features used are age, gender, country, ethnicity, education, neuroticism, openness to experience, extraversion, agreeableness, conscientiousness, impulsiveness, sensation seeking etc. Input features are given to the ANN-Dmodule to predict if volatile substance abuse (VSA) has been done by the person or not. ANN-C module predicts the use of VSA in terms of time such as day, week, month, year, decade, beforea decade, etc. The accuracy of the ANN-D module is found to be 81% and ANN-C module is 71.9%. Hence the proposed method can be used for drug analysis to some extent.