AoI-based Temporal Attention Graph Neural Network for Popularity Prediction and Content Caching

Abstract

Along with the fast development of network technology and the rapid growth of network equipment, the data throughput is sharply increasing. To handle the problem of backhaul bottleneck in cellular network and satisfy people's requirements about latency, the network architecture like informationcentric network (ICN) intends to proactively keep limited popular content at the edge of network based on predicted results. Meanwhile, the interactions between the content (e.g., deep neural network models, Wikipedia-alike knowledge base) and users could be regarded as a dynamic bipartite graph. In this paper, to maximize the cache hit rate, we leverage an effective dynamic graph neural network (DGNN) to jointly learn the structural and temporal patterns embedded in the bipartite graph. Furthermore, in order to have deeper insights into the dynamics within the evolving graph, we propose an age of information (AoI) based attention mechanism to extract valuable historical information while avoiding the problem of message staleness. Combining this aforementioned prediction model, we also develop a cache selection algorithm to make caching decisions in accordance with the prediction results. Extensive results demonstrate that our model can obtain a higher prediction accuracy than other stateof-the-art schemes in two real-world datasets. The results of hit rate further verify the superiority of the caching policy based on our proposed model over other traditional ways.

Semantics-enhanced Temporal Graph Networks for Content Popularity Prediction

“…Table II demonstrates the prediction performance of our proposed models (i.e., M1-STGN, M2-STGN), and their variants, as well as the original TGNs in [16], [17]. With a sparse dataset, it is clear that our models can yield better results in both transductive task and inductive task, even when we aggregate the semantic information only by summation.…”

“…1, it is intractable for classical GNN-based methods to predict the user preference in a sparse graph. But the attachment of semantics constructs more implicit structural patterns, which facilitates preference inference and may further improve the prediction performance of TGN models [17]. Additionally, a content might possess multiple genres (e.g., a fictional action movie containing both fiction and action genres) and the preferred genre might vary across users as well.…”

“…where the filtering and aggregation function Agg(•) can be implemented diversely. In the remainder of this paper, we take account of three means including filtering the latest message, using the mean value of all messages [16], and attention-based weighted summation of all messages [17]. We denote them as TGN-L, TGN-M and TGN-A, respectively.…”

“…The preprocessed dataset includes 86,889 interactions, which involve 11,254 different users and 4,057 pieces of content. Obviously, it is even much sparser than the dataset used in [17]. Afterwards, we perform a 60%-20%-20% chronological split of the dataset for training, validation, and testing, respectively.…”

“…Ref. [17] optimizes the temporal learning module of TGN with an age of information (AoI) based attention mechanism to filter and aggregate fresh historical messages, and realizes satisfactory content caching results.…”