Machine Learning for Disease Prediction Using Multi-Label Symptoms

Multi-label disease prediction is a highly significant task in healthcare, which essentially encompasses cases where several diseases may occur simultaneously, in a patient. The prediction accuracy suffers due to the fact that traditional models such as rule-based systems and decision trees together with Naïve Bayes fail to recognize symptom dependencies and imbalanced datasets. To this end, this study proposes a reliable machine learning system using the CatBoost, MLP, and Stacking Classifiers. On the same, the result obtained by the model Stacking Classifier is the best with an accuracy of 99.50%. This study greatly contributes to the development of multi-disease prediction systems: early detection, therapy preparation, and better outcomes.