Abstract:A major problem of making friend suggestions in social networks is the large size of social graphs, which can have hundreds of millions of people and tens of billions of connections. Classic methods based on heuristics or factorizations are often used to address the difficulties of scaling more complex models. However, the unsupervised nature of these methods can lead to suboptimal results. In this work, we introduce the Generalized Ego-network Friendship Score framework, which makes it possible to use complex supervised models without sacrificing scalability. The main principle of the framework is to reduce the problem of link prediction on a full graph to a series of low-scale tasks on ego-nets with subsequent aggregation of their results. Here, the underlying model takes an ego-net as input and produces a pairwise relevance matrix for its nodes. In addition, we develop the WalkGNN model which is capable of working effectively in the social network domain, where these graph-level link prediction tasks are heterogeneous, dynamic and featureless. To measure the accuracy of this model, we introduce the Ego-VK dataset that serves as an exact representation of the real-world problem that we are addressing. Offline experiments on the dataset show that our model outperforms all baseline methods, and a live A/B test demonstrates the growth of business metrics as a result of utilizing our approach.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the challenge of friend recommendation in social networks, especially the link prediction problem in large - scale social graphs. Specifically, social networks usually contain hundreds of millions of users and billions of connections, which makes it difficult for traditional methods to scale to such a large amount of data. In addition, although traditional heuristic - or factorization - based methods can handle large - scale data, their unsupervised nature may lead to sub - optimal results. To solve these problems, the author proposes a framework named "Generalized Ego - network Friendship Score". The core idea of this framework is to simplify the link prediction problem on the entire graph into a series of small - scale tasks on the ego - net (ego - centric network) and then aggregate the results of these tasks. The ego - net refers to a sub - graph centered on a certain node, including all its neighbors and the connections between them. Through this method, complex supervised models can be used without sacrificing scalability. To further improve the performance of the model, the author also develops a model named WalkGNN. WalkGNN is a second - order graph neural network (second - order GNN), which can effectively handle heterogeneous, dynamic and graph - level link prediction tasks without node features in the field of social networks. The key component of WalkGNN is the WalkConv layer, which regards each edge as an information filter and passes the state of the node pair through it. To verify the effectiveness of the model, the author introduces the Ego - VK dataset, which is a real - world dataset extracted from the VK social network and is used to evaluate the performance of the model in actual scenarios. The experimental results show that the WalkGNN model outperforms all baseline methods in offline experiments and also shows an increase in business metrics in online A/B tests. In summary, this paper aims to solve the challenges of friend recommendation in social networks and proposes a new framework and model that can improve the accuracy of link prediction while maintaining scalability.

GNN Applied to Ego-nets for Friend Suggestions

Self-Propagation Graph Neural Network for Recommendation

Friend Recommendations with Self-Rescaling Graph Neural Networks

A Graph-Neural-Network-Based Social Network Recommendation Algorithm Using High-Order Neighbor Information

Harnessing the Power of Ego Network Layers for Link Prediction in Online Social Networks

Ego-GNNs: Exploiting Ego Structures in Graph Neural Networks

Exploring implicit influence for social recommendation based on GNN

GNNRec: gated graph neural network for session-based social recommendation model

A Graph Neural Network Framework for Social Recommendations

Enhanced graph neural network for session-based recommendation

Graph neural networks for preference social recommendation

Enhanced Graph Neural Networks with Ego-Centric Spectral Subgraph Embeddings Augmentation

ApeGNN: Node-Wise Adaptive Aggregation in GNNs for Recommendation

Diversifying Collaborative Filtering via Graph Spreading Network and Selective Sampling

Xgcn: an Extreme Graph Convolutional Network for Large-scale Social Link Prediction

SocialGCN: an Efficient Graph Convolutional Network Based Model for Social Recommendation

Neighbor enhanced graph convolutional networks for node classification and recommendation

Generating Reliable Friends via Adversarial Training to Improve Social Recommendation

Incorporating A Triple Graph Neural Network with Multiple Implicit Feedback for Social Recommendation

GDSRec: Graph-Based Decentralized Collaborative Filtering for Social Recommendation

Heterogeneous Hypergraph Neural Network for Social Recommendation using Attention Network