Improving the Accuracy in Classification Using the Bayesian Relevance Feedback (BRF) Model

In this study, the components of a decision support system (DSS) are discussed. One of the main components is the inference engine (IE). In order to improve the IE in supporting the system's output, Bayesian Relevance Feedback (BRF) model is proposed. This BRF model as suggested has the potential to generate the most relevant target or classes (stage) based on relevance feedback knowledge. Subsequently, eight classifiers: Bayesian Model (BM), K-Nearest Neighbors (KNN), Meta MultiClass Classifier (MCC), Rule OneR (OneR), Random Tree (RT), Multilayer Perceptron (MLP), Naive Bayes (NB), and SMO-Poly Kernel (E-1.O) (SVM) are used in order to evaluate the efficiency of the proposed model. The empirical comparison shows that the BRF significantly improved the accuracy of the entire classification algorithm used for the oral cancer data set with a mean accuracy of 95.83%. It is also noted that the proposed model, BRF contributes to solving the posterior probabilities that do not exist (probability with zero values) in order to improve the decision-making in the oral cancer diagnosis.