Integrasi Deep Neural Network dan Rule-Based Reasoning dalam Sistem Pakar untuk Diagnosis Gangguan Sistem Saraf

Abstract

The advancement of artificial intelligence in healthcare has encouraged the use of deep learning to support medical diagnosis. However, Deep Neural Network (DNN) models suffer from low interpretability due to their black-box nature, which limits clinical applicability. This study aims to integrate DNN and Rule-Based Reasoning (RBR) into an expert system to provide accurate and explainable neurological disorder diagnosis. The dataset consists of 400 clinical patient records covering four diagnostic classes: peripheral neuropathy, transient ischemic attack (TIA), acute migraine, and epilepsy. The DNN model employs a multilayer perceptron architecture with two hidden layers and ReLU activation, while RBR applies IF–THEN rules derived from expert knowledge. The integration mechanism combines DNN probability and rule-based certainty factors through weighted scoring. Experimental results show an accuracy of 91%, precision of 0.90, recall of 0.89, and F1-score of 0.895. Expert validation indicates an 86% confidence level, demonstrating that the proposed system is suitable as an explainable artificial intelligence-based diagnostic support tool.