The rapid development of 5G and the widespread use of smart terminal devices bring explosive video traffic growth. Edge caching technology in mobile edge computing systems stores popular content of most interest to users in advance in edge servers closer to mobile users to alleviate the pressure of traffic congestion and excessive access latency caused by centralized storage. However, how to obtain popular videos while protect user data privacy with only implicit user feedback data, such as liking, viewing, favoriting, etc., is the key challenge. To tackle these challenges, we proposed a proactive video caching scheme based on federated learning and implicit feedback (FIPC). First, a mobile edge-cloud system model contained a three-tier network architecture is developed. Then, we introduced the federated process for training the denoised auto-encoder model and video caching in detail. Finally, experimental results in Movielens dataset show that, without user data leakage, the proposed FIPC scheme outperforms the baseline caching algorithm using user feedback data.

Spatial–temporal graph modeling plays an important role in the fields of transportation, meteorology, and social networks. Traffic flow prediction is a classic spatial–temporal modeling task. Existing methods usually do not take into account the asynchronous spatial–temporal correlation in traffic data. In addition, due to the complexity and variability of traffic data, long-term traffic forecasting is highly challenging. In order to solve the above problems, this article proposes a new deep learning-based asynchronous dilation graph convolution network (ADGCN) to model the spatial–temporal graphs. We mine the asynchronous spatial–temporal correlation in the traffic network, and propose the asynchronous spatial–temporal graph convolution (ASTGC) operation to extract this special relationship. Furthermore, we extend the dilated 1-D causal convolution to a graph convolution. The receptive field of the model increases exponentially with the increase of the network depth. Experiments are conducted on three public traffic data sets, and the results show that the prediction performance of ADGCN is better than the existing counterpart methods, especially in long-term prediction tasks.