An intelligent heart disease prediction system using hybrid deep dense Aquila network

Abstract

This paper proposes an intelligent HD prediction system based on a hybrid deep dense Aquila network for predicting HD at the early stage. The main intension of the proposed system is to provide an integrated deep learning model with advanced data mining approaches for framing effective decisions and accurate disease prediction. The different processing steps used in the proposed scheme are data acquisition, data pre-processing, optimized feature selection, and disease classification. Initially, the proposed system performs data acquisition to collect heart disease data from different public data sources. Then, the acquired data are pre-processed using noisy data elimination, mean computation and z-score or zero mean normalization. Next, an optimized feature selection model based on an enhanced sparrow search algorithm (E-SSA) is used to minimize the data dimensionality and selects the most optimal features from the pre-processed data. Finally, the selected optimal features are given as input to the deep-dense residual attention Aquila convolutional network (Deep-DenseAquilaNet) based disease classification model in which the weight is updated using the Aquila optimization algorithm (AOA). The proposed scheme is simulated in the Python platform and evaluated the performance in terms of different performance metrics using HD datasets (Statlog + Hungary + Cleveland + Switzerland + long beach VA datasets). Additionally, the performance of the proposed scheme is compared with recent existing algorithms. The maximum accuracy reached through the proposed scheme is 99.57%. Subsequently, the simulated results proved that the proposed scheme had achieved better performance than the existing schemes.