Tobacco Consumption and Mortality Rate Prediction using Machine Learning

Abstract

Cigarette smoking is one of the major reasons behind preventable deaths in the world and in India it was reported that approximately 28.6 percent of adults now smoked. Such a high level of prevalence poses a serious threat to the occurrence of chronic diseases and early death. The current research aims at forecasting the probability of a person dying as a result of using tobacco products by using machine learning methods incorporating lifestyle, medical, and demographic factors. Two models, Random Forest Regressor and Linear Regression were used to analyze a dataset of 2,000 patients. It was found that Random Forest was more predictive accurate (R 2 = 0.85, MAE = X ) than was Linear Regression (R 2 = 0.50, MAE = Y ). The analysis of feature importance also indicated that the risk was best determined by age, frequency of smoking, and access to healthcare services. These findings prove the existence of the potential of predictive analytics to help medical professionals to identify high-risk individuals in a more timely manner. Finally, such a strategy may be used to advocate the development of specific interventions and may help decrease the health toll of tobacco use in the long term.