Macro Graph Neural Networks for Online Billion-Scale Recommender Systems

TL;DR

This work tackles CTR prediction at billion-scale by addressing the computational and sampling biases of traditional GNNs. It introduces Macro Recommendation Graph (MAG) to group billions of micro nodes into hundreds of macro nodes and MacGNN to operate on this macro graph, enabling efficient online inference. Across four public and industrial datasets, MacGNN achieves superior offline performance, and online A/B tests on Taobao show improvements in PCTR, GMV, and engagement while maintaining fast latency. The approach provides a practical blueprint for scalable, graph-based CTR in large-scale recommender systems and points to macro-level graph representations as a promising direction for future research.

Abstract

Predicting Click-Through Rate (CTR) in billion-scale recommender systems poses a long-standing challenge for Graph Neural Networks (GNNs) due to the overwhelming computational complexity involved in aggregating billions of neighbors. To tackle this, GNN-based CTR models usually sample hundreds of neighbors out of the billions to facilitate efficient online recommendations. However, sampling only a small portion of neighbors results in a severe sampling bias and the failure to encompass the full spectrum of user or item behavioral patterns. To address this challenge, we name the conventional user-item recommendation graph as "micro recommendation graph" and introduce a more suitable MAcro Recommendation Graph (MAG) for billion-scale recommendations. MAG resolves the computational complexity problems in the infrastructure by reducing the node count from billions to hundreds. Specifically, MAG groups micro nodes (users and items) with similar behavior patterns to form macro nodes. Subsequently, we introduce tailored Macro Graph Neural Networks (MacGNN) to aggregate information on a macro level and revise the embeddings of macro nodes. MacGNN has already served Taobao's homepage feed for two months, providing recommendations for over one billion users. Extensive offline experiments on three public benchmark datasets and an industrial dataset present that MacGNN significantly outperforms twelve CTR baselines while remaining computationally efficient. Besides, online A/B tests confirm MacGNN's superiority in billion-scale recommender systems.

Figures (7)

• Figure 1: Illustration of neighbor number distributions in micro and macro user-item clicking interaction graphs within Taobao's billion-scale recommender system.
• Figure 2: Sketch map of the construction of the macro graph.
• Figure 3: The model architecture of the proposed MacGNN.
• Figure 4: The system architecture for online deployment.
• Figure 5: Efficiency study of the model inference time.