RETRACTED ARTICLE: A classification model to predict onset of smoking and drinking habits based on socio-economic and sociocultural factors

• Published in: Journal of ambient intelligence and humanized computing Vol. 12; no. 3; pp. 4171 - 4179
• Main Authors: Abirami, M. S., Vennila, B., Chilukalapalli, Eesh L., Kuriyedath, Rahul
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.03.2021; Springer Nature B.V
• Subjects: Addictive behaviors; Alcohol; Algorithms; Artificial Intelligence; Classifiers; Computational Intelligence; Drinking; Drug use; Economic analysis; Economic factors; Engineering; Habits; Machine learning; Mental disorders; Original Research; Peers; Regression analysis; Robotics and Automation; Smoking; Social factors; Sociocultural factors; Socioeconomic factors; Stress; Support vector machines; Teenagers; Tobacco; User Interfaces and Human Computer Interaction; Peer pressure; Deep learning; Logistic regression; Habits formation; Alcohol and other drugs (AODs); Machine learning; Socio-economic factors; Drinking; Socio-cultural factors; Smoking
• ISSN: 1868-5137; 1868-5145
• DOI: 10.1007/s12652-020-01796-4

Addictive habits are often initiated due to peer pressure. Smoking, drinking, and exercising are examples of such habits that are also performed as a part of socializing activities. Past researches have tried to predict the early onset of smoking and drinking problems using machine learning. However, these researches were mainly based on daily stress levels, and mood. Taking daily measures is inconvenient for a surveyor. The studies have also failed to account for socio-cultural and socio-economic factors which also play an important in the onset of these behaviours. Availability of items like tobacco and alcohol can significantly impact the onset of early alcohol and smoking in adolescents. In this paper, we analyze how socio-economic such as lifestyle, monthly savings, and socio-cultural factors like size of friends group, number of friends that drink and smoke, details about parents, etc., play a role in the initiation and cultivation of addictive behaviours and use a machine learning approach to predict the early onset of such behaviours. We compared Gaussian Naive Bayes, Support Vector Machine and Logistic Regression algorithms in order to train and predict our multi-classifier prediction system. We found Logistic Regression to be the best performing classifier to predict both drinking and smoking habits with 86.4% and 97.2% accuracies respectively. We also achieved an F1 measure of 0.76 for the drinking classifier and an F1 measure of 0.85 for the smoking classifier.