Context-based Fast Recommendation Strategy for Long User Behavior Sequence in Meituan Waimai

TL;DR

This paper tackles the challenge of long user behavior sequences in Meituan Waimai by introducing CoFARS, a context-based fast recommendation strategy that retrieves a sub-sequence aligned with contexts similar to the target context. It combines a probability encoder to map latent preferences to PoI-attribute distributions, a graph-based temporal aggregator to encode temporal dynamics, and a two-stage retrieval that uses prototypes to guide context matching. Offline experiments on a large Meituan Waimai dataset and online A/B tests show CoFARS outperforms baselines, with a CTR AUC improvement of 0.77% and a CTCVR AUC improvement of 0.63%, and online uplifts of 4.6% in CTR and 4.2% in GMV. The work demonstrates the value of leveraging contextual information in long-sequence modeling and has delivered production deployment benefits.

Abstract

In the recommender system of Meituan Waimai, we are dealing with ever-lengthening user behavior sequences, which pose an increasing challenge to modeling user preference effectively. Existing sequential recommendation models often fail to capture long-term dependencies or are too complex, complicating the fulfillment of Meituan Waimai's unique business needs. To better model user interests, we consider selecting relevant sub-sequences from users' extensive historical behaviors based on their preferences. In this specific scenario, we've noticed that the contexts in which users interact have a significant impact on their preferences. For this purpose, we introduce a novel method called Context-based Fast Recommendation Strategy to tackle the issue of long sequences. We first identify contexts that share similar user preferences with the target context and then locate the corresponding PoIs based on these identified contexts. This approach eliminates the necessity to select a sub-sequence for every candidate PoI, thereby avoiding high time complexity. Specifically, we implement a prototype-based approach to pinpoint contexts that mirror similar user preferences. To amplify accuracy and interpretability, we employ JS divergence of PoI attributes such as categories and prices as a measure of similarity between contexts. A temporal graph integrating both prototype and context nodes helps incorporate temporal information. We then identify appropriate prototypes considering both target contexts and short-term user preferences. Following this, we utilize contexts aligned with these prototypes to generate a sub-sequence, aimed at predicting CTR and CTCVR scores with target attention. Since its inception in 2023, this strategy has been adopted in Meituan Waimai's display recommender system, leading to a 4.6% surge in CTR and a 4.2% boost in GMV.

Figures (5)

• Figure 1: A User's behaviors vary in two characteristics. (a) Geographical location: user interests in different locations (such as companies, and homes) are quite different. (b) Dining time: preference of users at breakfast time (e.g. bun) is different from that at lunchtime (e.g. fast food) and midnight time (e.g. BBQ).
• Figure 2: An example of the JS divergence calculation based on the PoI attributes between different contexts.
• Figure 3: The overall architecture of Context-based User fast Recommendation Strategy($\textsc{CoFARS}\xspace$ for short) for long sequences in recommender system of Meituan Waimai.
• Figure 4: Performance variation of different prototype numbers
• Figure 5: Similarity between preferences under different contexts. The x-axis and y-axis both denote the contexts, $b$ stands for breakfast, $l$ stands for lunch, $d$ stands for dinner, $h$ represents home, and $c$ represents company.