Investigation of Machine Learning as Inference Engine of Expert System for Pharyngitis Diseases

Pharyngitis is an infectious disease that can be caused by bacteria or viruses as the type of pathogen that can pose diagnostic challenges due to its overlapping symptoms. Early warning of this disease is essential to reduce the risk of further transmission and fatality. Early warning using expert systems is one alternative, but traditional rule-based expert systems have difficulty in adapting to changes in reference data. Changes in potential disease data will develop over time which needs to be adjusted to the inference engine in the expert system. This study integrates machine learning (ML) as the inference engine of the expert system, with the aim of improving the diagnosis of bacterial and viral pharyngitis. The dataset consists of 716 clinical data examples, including pharyngitis symptoms, which are used for the purpose of training and testing the ML model. Several algorithms are tested, including Multi-Layer Perceptron (MLP), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), and Naïve Bayes (NB). The best performing algorithms, namely NB and RF, achieved accuracy values of around 97.54%, 98.04%, 92.59%, and 95.23%, respectively. This study shows that machine learning has the potential to support the improvement of the inference engine adaptability of expert systems, so it is expected to improve their diagnostic accuracy.