A Patient Network-Based Ensemble Machine Learning Framework for Chronic Diseases Prediction

The load on healthcare systems has increased due to the rise in the incidence of chronic diseases in the last decades. Understanding the course of chronic diseases and recognizing individuals at risk of emerging complications are critical, even though routine patient monitoring is costly and impractical. Artificial intelligence (AI) is consequently quickly changing the healthcare sector. Due to this, systems that rely on deep learning (DL) and machine learning (ML) are becoming more crucial to identifying and predicting diseases, whether based on medical data or photographs. By simulating human perception, these technologies provide crucial clinical support and have the potential to diagnose disorders that are challenging for people to detect. Current research satisfies this need by utilizing ML methods grounded in patient characteristics; nevertheless, they encounter challenges with high-dimensional data. To overcome these problems, we suggest a paradigm based on patient networks using Ensemble ML techniques for chronic illness prediction. We first project a patient-disease bipartite graph to develop a balanced patient network (BPN) that extracts the hidden characteristics among patients. We then construct prediction models using ML-based methods. These models generate strong patient representations for the prediction of chronic diseases by extracting network properties from BPN. The performance of many ensemble ML techniques was examined to ascertain the accuracy of chronic pulmonary illness and cardiovascular disease predictions. The outcomes demonstrate how our paradigm improves chronic illness prediction accuracy. The model predicts cardiovascular illness (CVD) with an accuracy of 93.49% and chronic pulmonary disease (CPD) with an accuracy of 89.15%. In addition, characteristics that were shown to contribute to the prediction included gender, age and the clustering coefficient. Healthcare professionals can benefit from this research by using their data more intelligently. Stakeholders can enhance healthcare delivery systems for individuals at risk of chronic illnesses and disorders by utilizing the suggested framework.