HiFIRec Towards High-Frequency yet Low-Intention Behaviors for Multi-Behavior Recommendation

TL;DR

HiFIRec tackles the challenge of heterogeneous multi-behavior data by explicitly distinguishing high-frequency, low-intention signals from more intentional interactions. It combines a graph-based embedding with a Hierarchical Noise Correction Module (layer-wise neighborhood aggregation and adaptive cross-layer fusion) and a Behavioral Intention Refinement Module, followed by an Intensity-Aware Non-Sampling strategy and joint multi-behavior optimization. Empirical results on Beibei and Taobao show state-of-the-art performance and robust improvements when removing or altering key components, validating the approach. The work advances robust multi-behavior recommendation by addressing both noisy signals and plausible but misleading frequent patterns, with practical implications for large-scale e-commerce and healthcare domains.

Abstract

Multi behavior recommendation leverages multiple types of user-item interactions to address data sparsity and cold-start issues,providing personalized services in domains such as healthcare and ecommerce.Most existing methods utilize graph neural networks to model user intention in a unified manner,which inadequately considers the heterogeneity across different behaviors.Especially,high frequency yet low intention behaviors may implicitly contain noisy signals,and frequent patterns that are plausible while misleading,thereby hindering the learning of user intentions.To this end,this paper proposes a novel multi-behavior recommendation method,HiFIRec,that corrects the effect of high-frequency yet low-intention behaviors by differential behavior modeling.To revise the noisy signals,we hierarchically suppress it across layers by extracting neighborhood information through layer-wise neighborhood aggregation and further capturing user intentions through adaptive cross layer feature fusion.To correct plausible frequent patterns,we propose an intensity-aware non-sampling strategy that dynamically adjusts the weights of negative samples.Extensive experiments on two benchmarks show that HiFIRec relatively improves HR@10 by 4.21%-6.81% over several state-of-the-art methods.

Figures (6)

• Figure 1: An illustration of high-frequency yet low-intention behaviors. High-frequency yet low-intention behaviors are the primary sources of noisy signals and plausible frequent patterns, to which the unified modeling of multi-behaviors is particularly prone. The Sub-figure (a) shows the noisy item "tea" may disturb intentions. Sub-figure (b) indicates the behaviors with different intention intensity may distort frequency-based distribution.
• Figure 2: The Overall Framework of HiFIRec. The framework primarily comprises four modules. (a) The Graph Embedding Module provides the representations for users, items, and behaviors. (b) The Hierarchical Noise Correction Module reduces noisy signals through Layer-Wise Neighborhood Aggregation and Adaptive Cross-Layer Feature Fusion. (c) The Behavioral Intention Refinement Module employs an adaptive behavior weighting to mitigate noisy signals effects. (d) The Multi-Behavior Prediction employs a non-sampling strategy to mitigate plausible frequent patterns.
• Figure 3: Datasets statistics
• Figure 4: The distribution of negative weight values for different types of behavior in the Taobao dataset and Beibei dataset.
• Figure 5: Comparison of negative weight performance of different types of behavior.