Approaches for handling imbalanced data used in machine learning in the healthcare field: A case study on Chagas disease database prediction

Abstract

Machine learning has increasingly gained prominence in the healthcare sector due to its ability to address various challenges. However, a significant issue remains unresolved in this field: the handling of imbalanced data. This process is crucial for ensuring the efficiency of algorithms that utilize classification techniques, which are commonly applied in risk management, monitoring, diagnosis, and prognosis of patient health. This study conducts a comparative analysis of techniques for handling imbalanced data and evaluates their effectiveness in combination with a set of classification algorithms, specifically focusing on stroke prediction. Additionally, a new approach based on Particle Swarm Optimization (PSO) and Naive Bayes was proposed. This approach was applied to the real problem of Chagas disease. The application of these techniques aims to improve the quality of life for individuals, reduce healthcare costs, and allocate available resources more efficiently, making it a preventive action.