Recommender for Its Purpose: Repeat and Exploration in Food Delivery Recommendations

TL;DR

This paper reveals that repeat and exploration behaviors in food delivery are shaped by distinct situational factors and should be modeled separately. It analyzes three real-world datasets to show historical situations predominantly affect repeat orders while collaborative cues enhance exploration. The authors propose two simple, situation-aware models, RepRec for repeat and ExpRec for exploration, plus an ensemble to merge outputs, and demonstrate superior performance over strong baselines on repeat, exploration, and combined tasks. The work underscores the importance of domain-specific analyses and lightweight, modular designs in recommender systems for practical, scenario-sensitive applications.

Abstract

Recommender systems have been widely used for various scenarios, such as e-commerce, news, and music, providing online contents to help and enrich users' daily life. Different scenarios hold distinct and unique characteristics, calling for domain-specific investigations and corresponding designed recommender systems. Therefore, in this paper, we focus on food delivery recommendations to unveil unique features in this domain, where users order food online and enjoy their meals shortly after delivery. We first conduct an in-depth analysis on food delivery datasets. The analysis shows that repeat orders are prevalent for both users and stores, and situations' differently influence repeat and exploration consumption in the food delivery recommender systems. Moreover, we revisit the ability of existing situation-aware methods for repeat and exploration recommendations respectively, and find them unable to effectively solve both tasks simultaneously. Based on the analysis and experiments, we have designed two separate recommendation models -- ReRec for repeat orders and ExpRec for exploration orders; both are simple in their design and computation. We conduct experiments on three real-world food delivery datasets, and our proposed models outperform various types of baselines on repeat, exploration, and combined recommendation tasks. This paper emphasizes the importance of dedicated analyses and methods for domain-specific characteristics for the recommender system studies.

Figures (5)

• Figure 1: Statistics on repeat consumption: (a) The percentages of orders being repeat at the $n$-th order of all users, (b) The number of explored stores at the $n$-th order of all users, (c) The cumulative distribution of repeat/exploration ratio in the last two weeks of users, and (d) that of all stores.
• Figure 2: Distributions of the influences of historical situations, $Inf_{his}$, on repeat and exploration consumption, where influences are quantified with the correlations between historical situation similarities and store similarities.
• Figure 3: Distributions of the influences of collaborative situations, $Inf_{col}$, on repeat and exploration consumption. Influences are quantified with the correlations between situation similarities and store similarities of interactions of collaborative users.
• Figure 4: The framework of our proposed recommenders considering both situations and repeat/exploration patterns for food delivery. Based on the observations made in Section \ref{['sec:dataAnalysis']}, RepRec and ExpRec are designed for repeat and exploration consumption, respectively. The Ensemble module combines their outputs if a unified recommendation list is needed.
• Figure 5: Performances of ExpRec and its variants without a specific trigger input. w/o is short for without.

Theorems & Definitions (2)

• Definition 1: Food Delivery Interaction
• Definition 2: Food Delivery Recommendation