Optimizing predictive accuracy: A comparative study of feature selection strategies in the healthcare domain

Abstract

Feature selection is a critical preprocessing step in the development of machine learning models, particularly in the healthcare domain, where datasets often contain numerous features that may not contribute significantly to predictive performance. This study presents a comparative analysis of various feature selection techniques applied to healthcare datasets, evaluating their effectiveness in improving model accuracy and reducing computational costs. We investigate traditional filter-based methods, such as information gain and chi-square, alongside wrapper-based approaches and hybrid techniques that combine the strengths of both. Using multiple healthcare datasets encompassing diverse medical conditions, we assess the impact of these techniques on classification performance using metrics such as accuracy, precision, recall, and F1-score. Additionally, we analyze the robustness and scalability of each method in handling high-dimensional data. The findings reveal significant differences in performance, highlighting the strengths and weaknesses of each feature selection approach within the healthcare context. This comparative study provides valuable insights for researchers and practitioners, guiding them in selecting appropriate feature selection techniques to enhance predictive modeling in healthcare applications.