Efficient utilization of novel feature selection strategies in various clinical applications

Abstract

Clinical data analysis using machine learning models has obtained newlineits importance in recent years due to its accurate predictions and classification newlineof samples, which supports clinicians greatly in the earlier diagnosis of newlinedisorders. A machine learning model is usually built with the base of data where newlinethe cleansing is carried out initially to remove the outliers. Once the newlinepreprocessed data is available, feature engineering plays a vital role in newlineperforming effective data analysis. Soon after calculating the better features, newlineclassification and prediction based on the data are carried out to diagnose newlinedisorders. Feature engineering here represents the feature extraction and newlineselection strategies performed to identify the key characteristic features that are newlinevery useful in representing the nature of the data. Extraction is computing raw newlinedata s features, including image, signal and text. newline Feature selection refers to choosing a set of important attributes from newlinethe available feature vector based on certain approaches. Three major newlineapproaches are widely used for feature selection: the filtering approach, the newlinewrapper approach and the embedded approach. The hybridization and the novel newlineapproach to feature selection are framed from these three major selection newlineapproaches. newline The efficiency of the feature selection and the classification methods newlinecan be analyzed based on their performance over the data. The proposed work newlinehas two novel feature selection approaches, namely Entropy Max Score newlineApproach (EMSA) and Random Forest Kerb (RFK) feature selection approach. newlineThese two approaches are framed to perform the real-time analysis of the newlineclinical data. newline