The Role of Mathematics in Machine Learning for Disease Prediction: An In-Depth Review in the Healthcare Domain

Abstract

The rapid advancements in healthcare technologies and the increasing complexity of medical data have made it imperative to explore and optimize predictive models for disease management. This study aims to conduct a systematic literature review to identify advancements, challenges, and opportunities in disease prediction using machine learning (ML) within the healthcare domain. The literature sources include Scopus, DOAJ, and Google Scholar, covering the period from 2013 to 2024. The findings reveal that both machine learning (ML) and deep learning (DL) algorithms have significant potential for disease prediction and treatment outcomes in various clinical contexts. Algorithms such as Random Forest, Logistic Regression, and ensemble techniques like Boosting have demonstrated strong performance in numerous studies. However, the effectiveness of these algorithms is highly context-dependent, including the type of disease, patient characteristics, and available data. Deep learning, particularly Convolutional Neural Networks (CNNs) and hybrid Long Short-Term Memory (LSTM) models, excels in handling complex, high-dimensional data, providing higher prediction accuracy compared to traditional ML models. This research shows that deep learning models, especially CNN and hybrid LSTM, achieve higher accuracy in disease prediction compared to traditional ML models. However, challenges related to data quality, privacy, and the underlying mathematical modeling of these algorithms remain to be overcome for wider applications.