Indian Journal of Science and Technology

Abstract

Objectives/Background: Addiction of alcohol is a complex disease which results from diversity of social, genetic and environmental influences. A report by World Health Organization, WHO (2014) estimates that most of the deaths are from alcohol related causes.The objective of this study is to analyze the alcoholic behavior of different age group people on the basis of risk factors. In this paper, we construct a comparative model of different classification techniques to analyze the best algorithm for predicting the alcoholic behavior of a person. Methods: Under this context, random tree and J48 that are decision tree algorithms have been exercised on the dataset of 600 people that is collected through a structured questionnaire by visiting de addicted centers, colleges, villages, government offices, old age homes of Patiala, Punjab. Findings: Results conclude that the random tree provides more precise results than J48 for all the age group people. Risk factors that come out to be most effective are impulsive nature, sensation seeking nature, financial loss, family conflict, depression, child abuse, alcoholic shop near home distance.The overall accuracy of random tree is 75.94% and for J48 is 71.26%. Applications/Improvement: There is a need to develop some intelligent tools in this area and the rules extracted from this analysis can be further used for designing the tool. More attributescan be incorporated to achieve the optimal results for predicting the behavior of an alcoholic person.
Keywords: Addiction, Classification, Data Mining, Prediction