A-ConvRNN: A Prediction Model for E-Commerce Page Views Based on Convolutional Neural Network and Attention Mechanism

Page view forecasting can provide an important basis for e-commerce website maintenance and merchant ordering. In this paper, we propose a recurrent neural network prediction model that incorporates convolutional neural network and attention mechanism. The model first computes the temporal features by recurrent neural network, then computes the key sequences in the temporal specialties by attention mechanism, and finally mines the temporal feature information by convolutional neural network, and outputs the prediction results using fully connected layers. By comparing with 11 time series baselines, the model achieves the best results for different prediction task lengths. We also validated the necessity of each component of the model through ablation studies.