Applying Explainable Machine Learning to Classify Smoking Status from Basic Health Biological Signals

Nowadays, the healthcare industry is undergoing a transformation with the increased availability of data. In particular, the data generated through visits of patients in their life span to the hospital known as electronic health records (EHR). This research presents a novel method for identifying cigarette smoking habits using measurements of fundamental health parameters, including heart rate, blood pressure, and oxygen saturation. Explainable machine learning techniques are applied to the top-performing machine learning algorithm to enhance human understanding of the results. The proposed approach, employing feature selection strategies and machine learning algorithms, accurately categorizes smoking habits. Evaluation using a publicly available dataset demonstrates an accuracy rate of 81 %. The model's interpretability is ensured by assessing the significance of input features and the model's decision-making process. This approach shows promise for clinical applications, facilitating early diagnosis of smoking-related health issues and the development of personalized smoking cessation programs. The methodology involves rigorous testing and validation, ensuring the reliability and robustness of the classification model. Further research could explore the application of this model in diverse populations and investigate the potential for integration with existing healthcare systems to improve public health outcomes.