A Novel Fuzzy Knowledge Graph Pairs-based Model in extreme cases for Hepatitis Sign Diagnosis

Abstract

Recently, machine learning models based on fuzzy inference systems have effectively solved problems in medical diagnosis and received much attention from researchers around the world. For example, Fuzzy Inference System (FIS), Mamdani Complex Fuzzy Inference System with Rule Reduction (M-CFIS-R), Mamdani Complex Fuzzy Inference System with Fuzzy Knowledge Graph (M-CFIS-FKG), and Fuzzy Knowledge Graph Pairs (FKG-Pairs) have been proposed. These machine-learning models have successfully solved most of the disease diagnosis problems based on complex symtom datasets with some characteristics such as amplitude and phase term, uncertain or incomplete input information. However, these models have revealed limitations when dealing with decision-making or disease diagnosis problems in extreme cases (where training datasets are too small or input datasets are large). To overcome this challenge, the FKG-Extreme model, based on FKG-Pairs combined with Q-learning techniques, has been proposed. In order to promote the practical application of the theoretical model FKG-Extreme, a model for Hepatitis Sign Diagnosis has been proposed in this paper. A simulation application of the proposed model has been built, and experimental results have been carried out based on the Liver dataset collected from the UCI machine learning repository, with results in terms of accuracy 13.3% higher than the FKG-Pairs model. Furthermore, some measures (Precision, Recall, and F1-Score) have also been used to evaluate performance and demonstrate that the proposed model has high reliability.