Leveraging Machine Learning for Early Detection and Treatment Decision Support in Oral Cancer

Oral cancer is emerging as an epidemic in South Asia. It is primarily associated with tobacco chewing, alcohol consumption, and betel quid, the most prevalent cultural practices of the area. It has long been emphasized by doctors that if the disease is detected early, patients tend to live longer. Early detection is also cheaper than treatment, a significant consideration for families and hospitals that are already heavily burdened. In this study we analyzed a large data set containing information of 85,000 oral cancer patients. This contains observations on various variables as age and gender, lifestyle, and early symptoms, as well as tumor size, stage, treatment, survival status, and issues of expense. To analyze and extract useful results from the data, different machine learning techniques, such as, logistic regression model, decision tree, random forest model, gradient boosting model, SVM algorithm, KNN algorithm, and Naïve Bayes algorithm have been employed. The numerical results depict that some variables i.e., the stage of disease, the size of the tumor, the cost of treatment, and survival time are more significant. The overall findings indicate a promising step toward technology-assisted diagnosis, offering new opportunities for healthcare systems facing resource limitations.