Oral Cancer Risk Factor Prediction Using Hybrid Fuzzy Logic and XGBoost: Experimental and Results

Oral cancer is a major cause of cancer morbidity and mortality globally and especially in developing countries. Late diagnosis greatly diminishes the chances of survival, and there is a necessity to use smart early detection tools. This paper suggests a combined predictive model that combines Fuzzy Logic and Extreme Gradient Boosting (XGBoost) in the determination of oral cancer risk based on both clinical and image-based predictors. Fuzzy logic is used to deal with uncertainty in the behavioral and demographic risk factors whereas XGBoost captures nonlinear relationships between the predictors. Through experimental analysis, the hybrid model has been proven to be superior to pure machine learning models. Accuracy, sensitivity, specificity, and AUC of the system were 91% and 90% and 92% respectively. The framework is highly interpretable and predictive, which is the reason it is applicable in a clinical decision support system or a screening application in a community context.