Discretize-Based Technique and Hybrid Machine Learning Approach for Medical Data Analysis and Mining

Abstract

With the extensive growth of data mining, its applications have expanded significantly, encompassing various fields, including healthcare. Despite the abundance of medical data, healthcare providers sometimes rely on personal observations rather than data-driven insights. To address this gap, researchers and medical professionals employ data mining techniques, notably machine learning algorithms, for health data analysis. In this study, three machine learning algorithms, namely K-nearest Neighbor, Random Forest, and Naive Bayes, are employed. Additionally, an ensemble majority voting approach is used in conjunction with a discretization data preprocessing technique, providing flexibility in model selection. Multiple datasets related to Heart disease, Breast tissue, Breast cancer, and Cryotherapy are utilized, offering a diverse range of data for analysis. Among the models tested, the discretized majority voting approach outperforms other classifiers and established state-of-the-art models in the literature. It achieves accuracy rates of 91.9%, 79.7%, 77.1%, and 88.9% for the Heart disease, Breast tissue, Breast cancer, and Cryotherapy datasets, respectively. This proposed methodology presents an effective and suitable learning approach for intelligent healthcare classification systems, especially when precision and model robustness are paramount.