Traffic Prediction Based on Random Connectivity in Deep Learning with Long Short-Term Memory
share
Traffic prediction plays an important role in evaluating the performance of telecommunication networks and attracts intense research interests. A significant number of algorithms and models have been proposed to learn knowledge from traffic data and improve the prediction accuracy. In the recent big data era, the relevant research enthusiasm remains and deep learning has been exploited to extract the useful information in depth. In particular, Long Short-Term Memory (LSTM), one kind of Recurrent Neural Network (RNN) schemes, has attracted significant attentions due to the long-range dependency embedded in the sequential traffic data. However, LSTM has considerable computational cost, which can not be tolerated in tasks with stringent latency requirement. In this paper, we propose a deep learning model based on LSTM, called Random Connectivity LSTM (RCLSTM). Compared to the conventional LSTM, RCLSTM achieves a significant breakthrough in the architecture formation of neural network, whose connectivity is determined in a stochastic manner rather than full connected. So, the neural network in RCLSTM can exhibit certain sparsity, which means many neural connections are absent (distinguished from the full connectivity) and thus the number of parameters to be trained is reduced and much fewer computations are required. We apply the RCLSTM solution to predict traffic and validate that the RCLSTM with even 35 shows a strong capability in traffic prediction. Also, along with increasing the number of training samples, the performance of RCLSTM becomes closer to the conventional LSTM. Moreover, the RCLSTM exhibits even superior prediction accuracy than the conventional LSTM when the length of input traffic sequences increases.
READ FULL TEXT
