END4Rec: Efficient Noise-Decoupling for Multi-Behavior Sequential Recommendation

TL;DR

END4Rec addresses the challenge of long, noisy multi-behavior sequences in sequential recommendation by integrating an Efficient Behavior Sequence Miner (EBM) with a dual denoising module and a Noise-Decoupling Contrastive Learning objective. EBM employs FFT-based frequency-domain fusion with a chunked diagonal mechanism and a compactness regularization to achieve $O(N \log N)$ complexity and compact parameter counts. The Hard Noise Eliminator and Soft Noise Filter tackle discrete and continuous noise signals, while the guided four-step training and contrastive loss promote robust decoupling of noise from signal. Experiments on three real-world datasets show END4Rec outperforms strong baselines in both effectiveness and efficiency, validating its practicality for real systems.

Abstract

In recommendation systems, users frequently engage in multiple types of behaviors, such as clicking, adding to a cart, and purchasing. However, with diversified behavior data, user behavior sequences will become very long in the short term, which brings challenges to the efficiency of the sequence recommendation model. Meanwhile, some behavior data will also bring inevitable noise to the modeling of user interests. To address the aforementioned issues, firstly, we develop the Efficient Behavior Sequence Miner (EBM) that efficiently captures intricate patterns in user behavior while maintaining low time complexity and parameter count. Secondly, we design hard and soft denoising modules for different noise types and fully explore the relationship between behaviors and noise. Finally, we introduce a contrastive loss function along with a guided training strategy to compare the valid information in the data with the noisy signal, and seamlessly integrate the two denoising processes to achieve a high degree of decoupling of the noisy signal. Sufficient experiments on real-world datasets demonstrate the effectiveness and efficiency of our approach in dealing with multi-behavior sequential recommendation.

Figures (5)

• Figure 1: Illustration of our motivations: increasing user interactions and the resulting large amount of data noise.
• Figure 2: Our END4Rec model, featuring an efficient user multi-behavior sequence mining process with $O(NlogN)$ complexity and adaptive input length. The framework consists of several key components: Efficient Behavior Sequence Miner (EBM) (4.3) for mining behavior patterns efficiently, Behavior-Aware Denoising (4.4) including Hard Noise Eliminator and Soft Noise Filter for noise detection, and Noise-Decoupling Contrastive Learning (4.5) for sufficient noise decoupling.
• Figure 3: Hyper-parameter study of the END4Rec.
• Figure 4: Visualization of hard noise removal ratios (left) and soft denoising kernels (right) for different behaviors.
• Figure 5: Comparison of Fourier spectral matrices with (left) and without (right) compactness regularization loss.