ABSTRACT
The Mel Frequency Cepstral Coefficient is designed to model features of audio signal and is widely utilized in various fields. Most of the researchers are used only 2–13 coefficients or 39 coefficients. MFCC selection count is an essential phase. Here MFCC selection phase is applied on heart disease prediction system. Heart produces different kinds of sounds such as normal and abnormal sounds if people affected by heart disease. First the heartbeat audio sounds are pre-processed. Then the feature extraction MFCC technique is applied on the heartbeat sounds. In this research proposes different kinds of MFCC filters are used with machine learning algorithms such as K-Nearest Neighbour, Support Vector Machine, Decision Tree,Random Forest, Naïve Bayes and Logistic Regression. Precision, recall, f1score and accuracy of the classifier with different count of MFCC features are used as the performance measures. This research finds MFCC count which provides maximum accuracy and better performance of mentioned six algorithms for identifying heart disease. The MFCC feature count depends on problem in different audio frequencies. Here selectively, Random Forest gives better accuracy with the MFCC feature of 40 for heart disease prediction system.
