Lung cancer risk prediction and feature importance analysis with machine learning algorithm

Abstract

The morbidity and mortality of lung cancer are high and the detection is difficult, which poses a great threat to people's health. Early detection and accurate diagnosis can significantly improve survival rates. In recent years, machine learning models have been widely applied to classify and predict the risk of getting lung cancer based on clinical features and certain environmental factors. In the research, a decision tree model and a random forest model are developed and validated to predict lung cancer risk using a dataset that contains patients living environments and clinical symptoms. The research also compared the amount of key features which are used in both decision tree and random forest model during prediction. Additionally, several key features in the prediction period are identified and are applied to a winform application designed for patients to test their risk level of getting lung cancer. The accuracy of the model used in winform application shows that the application can effectively predict lung cancer according to the clinical symptoms and living environment of patients, which proves that it has good application value.