Strategies for disease diagnosis by machine learning techniques

Machine learning (ML) techniques have become a point of interest in medical research. To predict the existence of a specified disease, two methods K-Nearest Neighbors (KNN) and logistic regression can be used, which are based on distance and probability, respectively. These methods have their problems, which leads us to use the ideas of both methods to improve the prediction of disease outcomes. For this sake, first, the data is transformed into another space based on logistic regression. Next, the features are weighted according to their importance in this space. Then, we introduce a new distance function to predict disease outcomes based on the neighborhood radius. Lastly, to decrease the CPU time, we present a partitioning criterion for the data.