Towards Popularity-Aware Recommendation: A Multi-Behavior Enhanced Framework with Orthogonality Constraint

TL;DR

This work tackles popularity bias in top-$K$ recommendations by leveraging multi-behavior side information through a tensor-based latent factor model (PopSI) and by explicitly factoring out item popularity via an orthogonality constraint. By modeling multiple user-item interactions as a 3D tensor with a shared latent space, PopSI improves the estimation of underlying preferences while mitigating bias. The orthogonal projection of item features away from popularity signals preserves expressive latent factors and reduces reliance on item popularity, achieving improved accuracy and debias performance with minimal tuning. Experiments on real-world e-commerce datasets demonstrate substantial gains over state-of-the-art debiasing methods, with scalable computation suitable for large-scale deployment.

Abstract

Top-$K$ recommendation involves inferring latent user preferences and generating personalized recommendations accordingly, which is now ubiquitous in various decision systems. Nonetheless, recommender systems usually suffer from severe \textit{popularity bias}, leading to the over-recommendation of popular items. Such a bias deviates from the central aim of reflecting user preference faithfully, compromising both customer satisfaction and retailer profits. Despite the prevalence, existing methods tackling popularity bias still have limitations due to the considerable accuracy-debias tradeoff and the sensitivity to extensive parameter selection, further exacerbated by the extreme sparsity in positive user-item interactions. In this paper, we present a \textbf{Pop}ularity-aware top-$K$ recommendation algorithm integrating multi-behavior \textbf{S}ide \textbf{I}nformation (PopSI), aiming to enhance recommendation accuracy and debias performance simultaneously. Specifically, by leveraging multiple user feedback that mirrors similar user preferences and formulating it as a three-dimensional tensor, PopSI can utilize all slices to capture the desiring user preferences effectively. Subsequently, we introduced a novel orthogonality constraint to refine the estimated item feature space, enforcing it to be invariant to item popularity features thereby addressing our model's sensitivity to popularity bias. Comprehensive experiments on real-world e-commerce datasets demonstrate the general improvements of PopSI over state-of-the-art debias methods with a marginal accuracy-debias tradeoff and scalability to practical applications. The source code for our algorithm and experiments is available at \url{https://github.com/Eason-sys/PopSI}.

Figures (7)

• Figure 1: The sources of popularity bias and the recommendation feedback loop.
• Figure 2: The sparsity of a sales transaction record subset from the Tmall dataset.
• Figure 3: A toy example of multiple feedback between users and items in e-commerce businesses.
• Figure 4: A toy example of matrix factorization scheme on the binary implicit feedback data.
• Figure 5: The illustration of the proposed popularity-aware recommendation algorithm PopSI.