HT-GNN: Hyper-Temporal Graph Neural Network for Customer Lifetime Value Prediction in Baidu Ads

2026-01-19

AI Summary

This work addresses two key challenges in predicting user lifetime value (LTV) for newsfeed advertising: the heterogeneity of LTV distributions across demographic segments and the irregular, rapidly evolving nature of user behavior sequences induced by dynamic marketing strategies. To tackle these issues, the authors propose HT-GNN, a novel model that integrates hypergraph neural networks with a Transformer-based temporal encoder. HT-GNN employs a hypergraph supervision module to capture cross-group relationships, an adaptive temporal weighting mechanism to model dynamic user interactions, and a task-adaptive mixture-of-experts (MoE) architecture to enable joint multi-horizon LTV prediction. Evaluated on a large-scale dataset of 15 million users from Baidu, HT-GNN consistently outperforms state-of-the-art methods across all evaluation metrics and prediction horizons.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Lifetime value (LTV) prediction is crucial for news feed advertising, enabling platforms to optimize bidding and budget allocation for long-term revenue growth. However, it faces two major challenges: (1) demographic-based targeting creates segment-specific LTV distributions with large value variations across user groups; and (2) dynamic marketing strategies generate irregular behavioral sequences where engagement patterns evolve rapidly. We propose a Hyper-Temporal Graph Neural Network (HT-GNN), which jointly models demographic heterogeneity and temporal dynamics through three key components: (i) a hypergraph-supervised module capturing inter-segment relationships; (ii) a transformer-based temporal encoder with adaptive weighting; and (iii) a task-adaptive mixture-of-experts with dynamic prediction towers for multi-horizon LTV forecasting. Experiments on \textit{Baidu Ads} with 15 million users demonstrate that HT-GNN consistently outperforms state-of-the-art methods across all metrics and prediction horizons.

Problem

• Customer Lifetime Value Prediction
• Demographic Heterogeneity
• Temporal Dynamics
• Irregular Behavioral Sequences
• News Feed Advertising

Innovation

• Hypergraph
• Temporal Dynamics
• Mixture-of-Experts
• Graph Neural Network
• Customer Lifetime Value