Cluster-based Graph Collaborative Filtering

TL;DR

This article proposes a novel GCN-based recommendation model, termed Cluster-based Graph Collaborative Filtering (ClusterGCF), which performs high-order graph convolution on cluster-specific graphs, which are constructed by capturing the multiple interests of users and identifying the common interests among them.

Abstract

Graph Convolution Networks (GCNs) have significantly succeeded in learning user and item representations for recommendation systems. The core of their efficacy is the ability to explicitly exploit the collaborative signals from both the first- and high-order neighboring nodes. However, most existing GCN-based methods overlook the multiple interests of users while performing high-order graph convolution. Thus, the noisy information from unreliable neighbor nodes (e.g., users with dissimilar interests) negatively impacts the representation learning of the target node. Additionally, conducting graph convolution operations without differentiating high-order neighbors suffers the over-smoothing issue when stacking more layers, resulting in performance degradation. In this paper, we aim to capture more valuable information from high-order neighboring nodes while avoiding noise for better representation learning of the target node. To achieve this goal, we propose a novel GCN-based recommendation model, termed Cluster-based Graph Collaborative Filtering (ClusterGCF). This model performs high-order graph convolution on cluster-specific graphs, which are constructed by capturing the multiple interests of users and identifying the common interests among them. Specifically, we design an unsupervised and optimizable soft node clustering approach to classify user and item nodes into multiple clusters. Based on the soft node clustering results and the topology of the user-item interaction graph, we assign the nodes with probabilities for different clusters to construct the cluster-specific graphs. To evaluate the effectiveness of ClusterGCF, we conducted extensive experiments on four publicly available datasets. Experimental results demonstrate that our model can significantly improve recommendation performance.

Figures (9)

• Figure 1: An example of user clustering. (a) A toy example containing 6 user nodes. (b) Hard clustering. Every user node can belong to only one cluster. (c) Soft clustering. Every user node may belong to several clusters with a fractional degree of membership.
• Figure 2: An overview of our ClusterGCF model with two cluster-specific graphs. In ClusterGCF, first-order propagation operates on the original user-item interaction graph, and high-order propagation operates on the cluster-specific graphs $\mathcal{G}_1$ and $\mathcal{G}_2$. The cluster-specific graphs are constructed by assigning probabilities to the nodes of the original user-item interaction graph based on their association with different clusters.
• Figure 3: An example of cluster-specific graphs construction. Each user and item node of the user-item interaction graph is classified into two clusters via a soft node clustering method. Specifically, each node in the original graph is allocated to the two clusters based on its soft assignment probabilities. Take user node $u_2$ as an example, it is assigned to cluster 1 with a weight of 0.2 and to cluster 2 with a weight of 0.8. Based on the soft assignment probabilities and the original edges from the user-item interaction graph, two cluster-specific graphs $\mathcal{G}_1$ and $\mathcal{G}_2$ are constructed.
• Figure 4: Performance comparison between ClusterGCF and Competitors at different layers on all datasets.
• Figure 5: Performance Comparison between ClusterGCF and IMP-GCN with different cluster numbers on all datasets.