A Novel Method for Predicting Time of Alcohol Use Based on Personality Traits and Demographic Information

Alcohol is considered to be a psychoactive substance that has dependence-producing properties. The early prediction of the degree of intake of alcohol can be helpful to get early intervention through counseling, and other preventive measures can be employed. Existing works related to alcohol use prediction have only focused on identifying whether an individual is an alcohol user or a non-user. There is no remarkable efficient model for predicting the time of use of alcohol to date. In this work, novel approaches have been proposed to predict the time of use of alcohol consumption by individuals. It predicts that the alcohol is consumed on just the previous day or previous week or previous month, or a previous year or not at all consumed. Ultimately, the risk of becoming an alcohol addict has been identified. The input features used in this study are demographic features and personality traits. SMOTE with Tomek link has been used to make the samples balanced in all the classes. Machine learning algorithms, such as K-nearest neighbour (k-NN), random forest and gradient boosting, have been used to predict the time of use based on input features. The highest accuracy of 77% is obtained in the random forest algorithm using the feature set comprising of demographic features and personality traits. The obtained sensitivity, specificity, precision and F-score are 76.198%, 94.19%, 75.53% and 75.86%, respectively. The novelty of the work is that it can predict the time of use of alcohol without a clinical test with acceptable accuracy.