Feature-Engineered Exploration: Prediction of Smoking Status using Machine Learning Ensemble Techniques and Survival Analysis Using Cox Proportional Hazard Model

This research delves into the complexities of smoking behavior by leveraging machine learning algorithms, to extract actionable insights from health data. By focusing on feature engineering methodologies, the research aims to identify key health metrics crucial for understanding smoking habits and predicting cessation outcomes. The prevalence of smoking remains a significant public health concern, necessitating a deeper understanding of the underlying dynamics driving smoking behavior. Traditional approaches to smoking status prediction and cessation estimation often lack precision and effectiveness, highlighting the need for innovative methods that can leverage large-scale health data to inform targeted intervention strategies. Utilizing existing machine learning techniques and integrating it with risk prediction, the accuracy for all tested models improved significantly. This research employs feature engineering to discern pivotal health metrics associated with smoking behavior. Through binary classification models, such as random forest and support vector machines, the research endeavors to accurately predict smoking status based on these identified features. Additionally, this research also utilizes the Cox proportional hazard model for precise estimation of smoking cessation timelines.