Abstract:Click-through Rate (CTR) prediction is crucial for online personalization platforms. Recent advancements have shown that modeling rich user behaviors can significantly improve the performance of CTR prediction. Current long-term user behavior modeling algorithms predominantly follow two cascading stages. The first stage retrieves subsequence related to the target item from the long-term behavior sequence, while the second stage models the relationship between the subsequence and the target item. Despite significant progress, these methods have two critical flaws. First, the retrieval query typically includes only target item information, limiting the ability to capture the user's diverse interests. Second, relational information, such as sequential and interactive information within the subsequence, is frequently overlooked. Therefore, it requires to be further mined to more accurately model user interests. To this end, we propose Multi-granularity Interest Retrieval and Refinement Network (MIRRN). Specifically, we first construct queries based on behaviors observed at different time scales to obtain subsequences, each capturing users' interest at various granularities. We then introduce an noval multi-head Fourier transformer to efficiently learn sequential and interactive information within the subsequences, leading to more accurate modeling of user interests. Finally, we employ multi-head target attention to adaptively assess the impact of these multi-granularity interests on the target item. Extensive experiments have demonstrated that MIRRN significantly outperforms state-of-the-art baselines. Furthermore, an A/B test shows that MIRRN increases the average number of listening songs by 1.32% and the average time of listening songs by 0.55% on a popular music streaming app. The implementation code is publicly available at <a class="link-external link-https" href="https://github.com/psycho-demon/MIRRN" rel="external noopener nofollow">this https URL</a>.

A new interest extraction method based on multi-head attention mechanism for CTR prediction

Click-Through Rate Prediction Algorithm Based on Modeling of Implicit High-Order Feature Importance

Click-through rate prediction model integrating user interest and multi-head attention mechanism

Time-aware Multi-layer Interest Extraction Network for Click-Through Rate Prediction

A Hierarchical Attention Model for CTR Prediction Based on User Interest

Multi-view Click-through Rate Prediction Based on Multi-layer Deep Interest Network

MISS: Multi-Interest Self-Supervised Learning Framework for Click-Through Rate Prediction

DisenCTR

Multi-Interactive Attention Network for Fine-grained Feature Learning in CTR Prediction

Deep Interest Evolution Network for Click-Through Rate Prediction

PC-IEN: a click-through rate prediction method based on dynamic collaborative personalized interest extraction

HIEN: Hierarchical Intention Embedding Network for Click-Through Rate Prediction

DisenCTR: Dynamic Graph-based Disentangled Representation for Click-Through Rate Prediction

Multi-granularity Interest Retrieval and Refinement Network for Long-Term User Behavior Modeling in CTR Prediction

Deep Context Interest Network for Click-Through Rate Prediction

DMBIN: A Dual Multi-behavior Interest Network for Click-Through Rate Prediction Via Contrastive Learning

Deep Group Interest Modeling of Full Lifelong User Behaviors for CTR Prediction

Practice on Effectively Extracting NLP Features for Click-Through Rate Prediction

Hierarchically Fusing Long and Short-Term User Interests for Click-Through Rate Prediction in Product Search

Evolving Interest with Feature Co-action Network for CTR Prediction