Session-Based Recommendation by Exploiting Substitutable and Complementary Relationships from Multi-behavior Data

TL;DR

This work tackles session-based recommendation by exploiting multi-behavior data (clicks and purchases) to uncover substitutable and complementary item relationships. It introduces SCRM, a graph-based model that constructs two undirected graphs, $ ext{G}^s$ and $ ext{G}^c$, to encode substitutes and complements from fused sessions, and applies a denoising layer to filter noise before learning item embeddings via weighted graph attention networks. A joint loss with three components, $L_r$, $L_{ex}$, and $L_{se}$, guides the model to predict the next item while enforcing exclusivity between relationships and encouraging semantic similarity among related items, with the final loss $L = L_r + eta_1 L_{ex} + eta_2 L_{se}$. Experiments on two real-world datasets show SCRM outperforms state-of-the-art baselines, with ablations validating the importance of the denoising layer, the separate graphs, and the loss terms. The results demonstrate that explicitly modeling substitutable and complementary relationships from multi-typed behaviors yields meaningful improvements for SR in e-commerce settings and offers a framework for integrating additional behavior types in the future.

Abstract

Session-based recommendation (SR) aims to dynamically recommend items to a user based on a sequence of the most recent user-item interactions. Most existing studies on SR adopt advanced deep learning methods. However, the majority only consider a special behavior type (e.g., click), while those few considering multi-typed behaviors ignore to take full advantage of the relationships between products (items). In this case, the paper proposes a novel approach, called Substitutable and Complementary Relationships from Multi-behavior Data (denoted as SCRM) to better explore the relationships between products for effective recommendation. Specifically, we firstly construct substitutable and complementary graphs based on a user's sequential behaviors in every session by jointly considering `click' and `purchase' behaviors. We then design a denoising network to remove false relationships, and further consider constraints on the two relationships via a particularly designed loss function. Extensive experiments on two e-commerce datasets demonstrate the superiority of our model over state-of-the-art methods, and the effectiveness of every component in SCRM.

Figures (11)

• Figure 1: Examples showcasing users' historical interactions.
• Figure 2: The process of graph construction.
• Figure 3: The overview of our proposed SCRM model.
• Figure 4: Denoising network.
• Figure 5: Integrating substitutable and complementary graphs.

Theorems & Definitions (1)

• Definition 1