On the Practice of Deep Hierarchical Ensemble Network for Ad Conversion Rate Prediction

2025-04-10

AI Summary

In conversion rate (CVR) prediction for performance advertising, key challenges include extreme label sparsity, fragmented modeling of on-platform and off-platform user behaviors, and heavy reliance on heuristic feature interaction design. To address these, this paper proposes a multi-task CVR joint modeling framework based on a Deep Hierarchical Ensemble Network (DHEN). Methodologically: (1) a unified DHEN backbone integrates real-time on-platform behavioral sequences with off-platform conversion sequences; (2) it introduces the first dual-sequence modeling paradigm explicitly capturing both on- and off-platform dynamics; (3) a self-supervised auxiliary task—future action prediction over behavioral sequences—is incorporated to mitigate label sparsity. The work systematically investigates cross-module composition strategies (e.g., Transformer, DCN, MLP), depth-width trade-offs, hyperparameter configurations, and pretrains personalized user representations. Evaluated on large-scale industrial advertising data, DHEN achieves state-of-the-art performance, significantly outperforming single-module baselines.

Technology Category

• Computer Vision: Multi-modal Vision
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Machine Learning: Ensemble Methods

Application Category

• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

The predictions of click through rate (CTR) and conversion rate (CVR) play a crucial role in the success of ad-recommendation systems. A Deep Hierarchical Ensemble Network (DHEN) has been proposed to integrate multiple feature crossing modules and has achieved great success in CTR prediction. However, its performance for CVR prediction is unclear in the conversion ads setting, where an ad bids for the probability of a user's off-site actions on a third party website or app, including purchase, add to cart, sign up, etc. A few challenges in DHEN: 1) What feature-crossing modules (MLP, DCN, Transformer, to name a few) should be included in DHEN? 2) How deep and wide should DHEN be to achieve the best trade-off between efficiency and efficacy? 3) What hyper-parameters to choose in each feature-crossing module? Orthogonal to the model architecture, the input personalization features also significantly impact model performance with a high degree of freedom. In this paper, we attack this problem and present our contributions biased to the applied data science side, including: First, we propose a multitask learning framework with DHEN as the single backbone model architecture to predict all CVR tasks, with a detailed study on how to make DHEN work effectively in practice; Second, we build both on-site real-time user behavior sequences and off-site conversion event sequences for CVR prediction purposes, and conduct ablation study on its importance; Last but not least, we propose a self-supervised auxiliary loss to predict future actions in the input sequence, to help resolve the label sparseness issue in CVR prediction. Our method achieves state-of-the-art performance compared to previous single feature crossing modules with pre-trained user personalization features.

Problem

• Optimizing Deep Hierarchical Ensemble Network for ad conversion rate prediction
• Determining optimal feature-crossing modules and hyper-parameters in DHEN
• Addressing label sparseness in CVR prediction with self-supervised learning

Innovation

• Multitask learning with DHEN backbone
• On-site and off-site behavior sequences
• Self-supervised auxiliary loss for sparsity