Study of Wavelet and Line Search Techniques on Modified Backpropagation Polak-Ribiere Algorithm for Heart Failure Detection

Congestive Heart Failure (CHF) is a disease due to abnormalities in heart muscles so the heart not able to pump the bloods according to the body needs. Heart signals can be detected using Electrocardiography (ECG). However, there are no specific ECG features of CHF patients, whereas the extracted features of ECG signals play a significant role for diagnosing the cardiac disease. In this paper, we used Discrete Wavelet Transform (DWT) and Wavelet Package Decomposition (DWT) to extract the features. As for the process of this work is divided into three phases, i.e. pre-processing, feature extraction, and classification. Thus, the extracted features will then be used as inputs for the classification system we used; Artificial Neural Network (ANN) Modified Backpropagation (MBP) Polak-Ribiere Conjugate Gradient with line search technique. At the end of the study, the feature was obtained using WPD at 5th level with 22 records of training data. Gained an average value that is higher than the other trials, 72.5%. For the classification, known that 30 neurons in hidden layer and Charalambous' Search is the fastest search technique to be applied to this case with processing time 2.65 seconds, 14 epochs, and 87.5% accuracy.