An enhanced machine learning approach for effective prediction of IgA nephropathy patients with severe proteinuria based on clinical data

IgA Nephropathy (IgAN) is a disease of the glomeruli that may eventually lead to chronic kidney disease or kidney failure. The signs and symptoms of IgAN nephropathy are usually not specific enough and are similar to those of other glomerular or inflammatory diseases. This makes a correct diagnosis more difficult. This study collected data from a sample of adult patients diagnosed with primary IgAN at the First Affiliated Hospital of Wenzhou Medical University, with proteinuria ≥1 g/d at the time of diagnosis. Based on these samples, we propose a machine learning framework based on weIghted meaN oF vectOrs (INFO). An enhanced COINFO algorithm is proposed by merging INFO, Cauchy Mutation (CM) and Oppositional Based Learning (OBL) strategies. At the same time, COINFO and Support Vector Machine (SVM) were integrated to construct the BCOINFO-SVM framework for IgAN diagnosis and prediction. First, the proposed enhanced COINFO is tested on the IEEE CEC2017 benchmark problems, and the results prove its efficient optimization ability and convergence accuracy. Furthermore, the feature selection capability of the proposed method is verified on the UCI public medical datasets. Finally, the auxiliary diagnostic experiment was carried out through IgAN real sample data. The results demonstrate that the proposed BCOINFO-SVM can screen out essential features such as High-Density Lipoprotein (HDL), Uric Acid (UA), Cardiovascular Disease (CVD), Hypertension and Diabetes. At the same time, the accuracy of the BCOINFO-SVM model can reach 98.56%, the sensitivity reaches 96.08%, and the specificity reaches 97.73%. It can become a potential auxiliary diagnostic model of IgAN.