Disease Diagnosis Using Multi Layer Perceptron

Abstract

The rapid growth of artificial intelligence in healthcare has enabled the development of intelligent systems that support early disease diagnosis and patient guidance. This paper presents an extended version of an intelligent healthcare virtual assistant that performs symptom-based disease diagnosis using Natural Language Processing (NLP) and a Convolutional Neural Network (CNN). The proposed system allows users to interact through a virtual assistant interface, where textual symptom inputs are processed and transformed into structured features for deep learning–based disease prediction. The CNN model effectively learns complex symptom–disease relationships and achieves an accuracy of 95.02%, with 98.44% precision, 95.14% recall, and an F1-score of 96.33%, demonstrating superior performance compared to traditional machine learning approaches. In addition to disease prediction, the extended system provides diet recommendations and relevant doctor or hospital information to assist users in making informed healthcare decisions. Although the system does not replace professional medical consultation or support real-time appointment booking, it serves as a reliable preliminary decision-support tool. The experimental results validate the effectiveness, accuracy, and practical applicability of the proposed CNN-based virtual assistant, highlighting its potential to reduce time, cost, and dependency on frequent hospital visits while improving accessibility to healthcare services.