Study And Analysis Of Feature Selection Techniques For Predicting Cardiovascular Diseases

Abstract

newline vi newlineABSTRACT newlineCardiovascular Disease (CVD) is the leading cause of morbidity and mortality newlineworldwide. Over three-quarters of CVD deaths take place in developing countries. newlineTherefore, it is essential to detect cardiovascular disease as early as possible to newlineprevent death. However, disease analysis and prediction are highly formidable in newlinemedical data analysis. newlineRecent advancements in data mining and the need for automated models have newlinepaved the way for developing a more reliable and efficient model for predicting CVD. newlineThe amount of data in the healthcare industry is vast. Data mining turns this large newlinecollection of raw healthcare data into information that can help to make informed newlinedecisions and predictions. Several kinds of research have been carried out in newlinepredicting CVD, but the focus on choosing the important attributes that play a newlinesignificant role in predicting CVD is inadequate. Hence, choosing the right features newlinefor the classification and diagnosis is important. newlineThis research discusses the challenges of selecting relevant features in the newlineproblem specification. This research aims to identify and select important features and newlinemachine learning methodologies that can enhance the prediction capability of the newlineclassification models for accurately predicting CVD. This research proposes two new newlinefeature selection methods, namely HRFLC and ASBA, to select significant features, newlinewhich aids in predicting the diseases accurately. The research uses the UCI dataset; newlinethe first method HRFLC identifies 10 significant features with the Naive Bayes newlinealgorithm producing higher accuracy by implementing various machine learning newlinealgorithms. The second method ASBA identifies 9 significant features with Gradient newlineBoosting algorithm producing higher accuracy than other machine learning newlinealgorithms. The results show that the proposed HRFLC and ASBA technique newlineoutperform the existing approaches in terms of precision, recall, F1score and newlineaccuracy. The experimental outcomes show that the proposed approach attains the newlinemaximum classificat