Deep Recurrent Encoder Network and Spark Model for Angiographic Disease Risk Classification

Abstract

Analysis and extraction of effective results from the medical big data is very complex due to the existence of large volume of data and it is also very difficult to classify the risk of angiographic diseases. The challenges faced by the conventional methods are complexity in classifying the data and performance degradation for a large-sized dataset. Hence, a hybrid approach named Deep Recurrent Encoder Network and Spark Model for Angiographic Disease Risk Classification (DRE-NET) is developed in this research to achieve accurate results for classifying the risk of angiographic diseases using Spark architecture. The developed Adaptive RCOA integrates Rider Optimization Algorithm (ROA) and Chicken Swarm Optimization (CSO) with the adaptive concept. The disease risk classification process is obtained by the Recurrent Neural Network (RNN) and Deep Stacked Autoencoder (DSAE) in such a way that the weights are optimally trained by the developed Adaptive RCOA. The optimal solution is obtained by evaluating the fitness function such that the fitness with minimal error value is considered the best solution. Moreover, the developed Adaptive RCOA-based RNN+DSAE attained better result with the metrics such as specificity, accuracy, and sensitivity with the values of 100%, 97.69%, and 99.19%, respectively, using the KDD Cup dataset.