Real-Time Symptomatic Disease Predictor Using Multi-Layer Perceptron

Abstract

Early disease diagnosis is crucial for effective treatment, but current healthcare methods have limitations. Supervised machine learning algorithms, particularly deep learning networks, have proven effective in developing medical diagnostics and real-time applications for detecting high-risk diseases. This paper evaluates five algorithms: Multilayer perceptron (MLP), random forest, decision tree, Naive Bayes, and K-Nearest neighbours (KNN) for predicting diseases based on user-entered symptoms. MLP outperformed other algorithms, achieving an accuracy of 97.2%, which is 4-5% higher than existing disease prediction models. Notably, existing techniques account for only 94% accuracy on average. Highlighting the potential of MLP in early disease diagnosis, this paper concludes by summarizing its goals, challenges, and outcomes.