BiLSTM_SAE:A Hybrid Deep Learning Framework for Efficient Predictive Big Data Analytics System

Abstract

Abstract Big data has been utilized and attracted various researchers due to the phenomenal increase in computational application which has developed an overwhelming flow of data. Further, with an expeditious blooming of emerging applications such as social media applications, semantic Web, and bioinformatics applications, data heterogeneity is increasing swiftly. Accordingly, a variety of data needs to be executed with less high accuracy and less. However, effective data analysis and processing of large-scale data are compelling which is considered a critical challenge in the current scenario. To overcome these issues, various techniques have been developed and executed but still, it is significant to improve in accuracy. The current study proposed a hybrid technique of BiLSTM-SAE has been proposed for business big data analytics. Bidirectional LSTM is an advanced version of the conventional LSTM approach. The performance comparison of the proposed method BiLSTM-SAE with existing Random forest-RF has been processed. The final result reported that the proposed method BiLSTM-SAE had been procured with better accuracy of 0.836. Moreover, the training and validation accuracy and loss on different performance metrics have been conducted and studied in the research.