Multi-Disease Prediction Method Based on Machine Learning and Computational Approaches

Abstract

In today's era, various life-threatening illnesses such as those related to the heart,  diabetes, hypothyroid, breast cancer, and others are rapidly increasing the rate of sudden  deaths. Early detection is crucial for saving lives and preventing severe illnesses. While  existing research has shown promising forecasting rates ranging from 75% to 92%, there  remains an accuracy gap of 8% to 25% to be addressed. This paper introduces a novel approach  to multi-disease forecasting, aiming to significantly improve prediction accuracy through the  collaborative application of machine learning (ML) techniques. Our proposed method analyses the predictive capabilities of five different ML algorithms: Hidden Markov Model (HMM),  Support Vector Machine (SVM), DTNB (an ensemble algorithm combining Naïve Bayes and  decision trees), Artificial Neural Network (ANN), and Radial Basis Function Network (RBNF). By leveraging the complementary strengths of these algorithms, our approach  mechanically combines the forecasting abilities of top two well matched algorithms into a  layered structure, enabling data analysis at multiple stages. To evaluate our method, we utilize  diverse disease datasets sourced from the well-established UCI benchmarked open access  research repository. The comparative experimental results demonstrate the effectiveness of  our proposed approach, indicating significant improvements in disease forecasting accuracy.