A decision-making system for medical diagnosis based on iterative nearest component analysis and optimized learning

Abstract

A silent disease is a chronic illness without clear clinical symptoms, and is diagnosed at an advanced stage when there is irreversible damage. Increasingly, automatic machine methods are used for early diagnosis to reduce complications associated with diseases. The performance of previous automated methods, however, was plagued by uncertainty, lack of generalizability, and unreliability. Using the feature aggregation approach and optimized learning, this paper proposes a hybrid model incorporating iterative neighborhood component analysis (iNCA). Using the support vector machine (SVM) algorithm, which has been optimized from the water cycle algorithm (WCA) in the direction of classification, the best results have been reported from the classification of several types of diseases. A key feature of the WCA algorithm is its ability to find the global optimum. When selecting features, the method works rapidly and selects the subset of features that has the lowest error level. In this research, accuracy of diagnostics will be improved and the effects of overfitting will be reduced. We obtained the desired medical data from the UCI database, which contains diseases such as hepatitis, diabetes, kidney failure, and breast cancer datasets. As compared to similar methods that have been published in the last few years for automatic detection of silent diseases, it can be predicted that the results will be better.