An efficient Multi-Layer Hybrid Neural Network and optimized parameter enhancing approach for traffic prediction in Big Data Domain

Abstract

Network Traffic Prediction and analysis is one of the essential tasks for cellular networks to reduce the network load. Moreover, traffic data are categorized as “Big Data” for developing a prediction framework to achieve higher metrics. Furthermore, the recent adoption of deep learning has been trending research, concerning cross-domain like Big Data and cellular networks; however, there are still several gaps considering the higher prediction rate and minimal error rate. This research work develops Multi-Layer Hybrid Network (MLHN) for network traffic prediction and analysis; MLHN comprises the three distinctive networks for handling the different inputs for custom feature extraction. The first network takes the network traffic sequence as the input for a given specific period. The second network takes input as the references along with data that corresponds to the date and time of traffic and the third input take the cross-domain parameter for understanding and exploiting the deep features. Furthermore, an optimized and efficient parameter-tuning algorithm is introduced to enhance parameter learning which tends to reduce error. MLHN is evaluated considering the “Big Data Challenge” dataset considering the Mean Absolute Error, Root Mean Square Error and R^2  as metrics; furthermore, MLHN efficiency is proved through comparison with state-of-art approach.