Towards Smart Diagnostic: A Multi-Disease Prediction System Leveraging Machine Learning Techniques

Technological advancements in machine learning have revolutionized healthcare, enabling more accurate diagnoses and successful treatment of chronic diseases. Most existing machine-learning models focus on predicting or detecting a single disease. This study proposes a robust multi-disease prediction system based on machine learning to facilitate early and reliable diagnosis. Instead of focusing on a limited number of diseases, our approach utilizes a single dataset encompassing 41 unique diseases, eliminating the need for separate models. Unlike conventional methods that apply dimensionality reduction techniques like PCA, we preserve crucial features through correlation-based selection. An ensemble model like soft voting (SVM, DT, RF, NB) and a hybrid model is multitask learning (MTL) to enhance predictive accuracy. Moreover, Adaboost and CNN are also implemented to test their performance on the dataset. The system benefits from the ability to predict multiple diseases simultaneously, improving efficiency and reducing the complexity of disease classification tasks. Moreover, the integration of MTL optimizes the learning process by capturing shared information between diseases, resulting in more accurate predictions across diverse healthcare conditions. The proposed system achieves superior performance, with MTL recording an accuracy of 99%. The model provides a scalable solution for early disease detection, enhancing diagnostic accuracy and patient care.