An Explainable AI Approach to Classify Smoker and Drinker Profiles via Medical Parameters

Smoking and consumption of excess alcohol is one of the major rising concerns in our society. In accordance with a WHO report released on July 31, 2023, tobacco use kills over 8 million people annually, of which 1.3 million are non-smokers who are exposed to passive smoking, and three million die from alcohol abuse. It not only affects the physical and mental health of the people who practices it but also affects the people who are around them. This work will play an important role as it helps in identifying these unhealthy habits from medical parameters which can contribute in prevention of the health risks which lie ahead. The results obtained are beneficial for the intended users like health professionals, insurance policy handlers and public health agencies where they will be able to identify the people who show addiction to these habits. We have performed various machine learning classification models which includes Light Gradient Boosting (LGB), Logistic Regression, Extreme Gradient Boosting (XGB), Random Forest, K-Nearest Neighbor etc, on the dataset which contains medical parameters as features which is used to categorize a person as a smoker or not a smoker or someone who used to smoke, and a drinker or not. Also, the explainability tools used helps to understand the key features which contribute to the decision making in the classification