Development of machine learning models for post-operative recurrence prediction in lung cancer patients

Abstract

Surgical treatment is one of the best approaches to provide a better cure for lung cancer patients. Despite the technological advancements, the increase in lung cancer recurrence rate urges the development of an early-stage predictive model. Therefore, we carried out machine learning algorithms to predict post-operative recurrence in lung cancer patients. It is to note that 80% of patient data was used for the model development and 20% of patient data was used for validation of the model. Besides, the important parameters were found using the extra tree classifier and correlation analysis. Notably, OS, DFS time and tumor size were ensured higher importance during the feature selection process. Random forest achieved the highest accuracy score of 96% than the other algorithms investigated in this study. Indeed, prior consideration of the important features together with the random forest algorithm will help surgeons to make effective treatment progress in lung cancer patients.