Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Repeat-bias-aware Optimization of Beyond-accuracy Metrics for Next Basket Recommendation

Yuanna Liu, Ming Li, Mohammad Aliannejadi, Maarten de Rijke

TL;DR

The paper tackles repeat bias in next basket recommendation (NBR), where repeated items dominate utility and hinder beyond-accuracy objectives like diversity and item fairness. It introduces two MILP-based, repeat-bias-aware re-ranking algorithms, RADiv (diversity) and RAIF (item fairness), to post-process top-N candidates and balance relevance, beyond-accuracy metrics, and repeat ratio for both unified and combined NBR methods. Empirical evaluation on three grocery datasets shows that RADiv and RAIF can significantly improve diversity and fairness while mitigating repeat bias with only modest Recall loss, validating the practicality of re-ranking to harmonize utility and long-term objectives. The work highlights the importance of jointly optimizing repeat bias with beyond-accuracy metrics and points to future work in formalizing repeat bias measurement and exploring its definition across settings.

Abstract

In next basket recommendation (NBR) a set of items is recommended to users based on their historical basket sequences. In many domains, the recommended baskets consist of both repeat items and explore items. Some state-of-the-art NBR methods are heavily biased to recommend repeat items so as to maximize utility. The evaluation and optimization of beyond-accuracy objectives for NBR, such as item fairness and diversity, has attracted increasing attention. How can such beyond-accuracy objectives be pursued in the presence of heavy repeat bias? We find that only optimizing diversity or item fairness without considering repeat bias may cause NBR algorithms to recommend more repeat items. To solve this problem, we propose a model-agnostic repeat-bias-aware optimization algorithm to post-process the recommended results obtained from NBR methods with the objective of mitigating repeat bias when optimizing diversity or item fairness. We consider multiple variations of our optimization algorithm to cater to multiple NBR methods. Experiments on three real-world grocery shopping datasets show that the proposed algorithms can effectively improve diversity and item fairness, and mitigate repeat bias at acceptable Recall loss.

Repeat-bias-aware Optimization of Beyond-accuracy Metrics for Next Basket Recommendation

TL;DR

The paper tackles repeat bias in next basket recommendation (NBR), where repeated items dominate utility and hinder beyond-accuracy objectives like diversity and item fairness. It introduces two MILP-based, repeat-bias-aware re-ranking algorithms, RADiv (diversity) and RAIF (item fairness), to post-process top-N candidates and balance relevance, beyond-accuracy metrics, and repeat ratio for both unified and combined NBR methods. Empirical evaluation on three grocery datasets shows that RADiv and RAIF can significantly improve diversity and fairness while mitigating repeat bias with only modest Recall loss, validating the practicality of re-ranking to harmonize utility and long-term objectives. The work highlights the importance of jointly optimizing repeat bias with beyond-accuracy metrics and points to future work in formalizing repeat bias measurement and exploring its definition across settings.

Abstract

In next basket recommendation (NBR) a set of items is recommended to users based on their historical basket sequences. In many domains, the recommended baskets consist of both repeat items and explore items. Some state-of-the-art NBR methods are heavily biased to recommend repeat items so as to maximize utility. The evaluation and optimization of beyond-accuracy objectives for NBR, such as item fairness and diversity, has attracted increasing attention. How can such beyond-accuracy objectives be pursued in the presence of heavy repeat bias? We find that only optimizing diversity or item fairness without considering repeat bias may cause NBR algorithms to recommend more repeat items. To solve this problem, we propose a model-agnostic repeat-bias-aware optimization algorithm to post-process the recommended results obtained from NBR methods with the objective of mitigating repeat bias when optimizing diversity or item fairness. We consider multiple variations of our optimization algorithm to cater to multiple NBR methods. Experiments on three real-world grocery shopping datasets show that the proposed algorithms can effectively improve diversity and item fairness, and mitigate repeat bias at acceptable Recall loss.
Paper Structure (11 sections, 7 equations, 2 figures, 3 tables, 1 algorithm)

This paper contains 11 sections, 7 equations, 2 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Problem formulation
  5. Model overview
  6. Optimization objectives
  7. Repeat-bias-aware optimization for unified NBR methods
  8. Repeat-bias-aware optimization for combined NBR methods
  9. Experimental Setup
  10. Experimental Results
  11. Conclusion

Figures (2)

  • Figure 1: Performance of item fairness and diversity optimization without considering $RepRatio_{rec}$ on different NBR methods. logDP$\circ$ means the closer logDP value is to zero, the fairer it is.
  • Figure 2: Trade-off between Recall and diversity, item fairness, repeat bias. Take UP-CF@r as an example on the Instacart dataset. The red star indicates the optimal solution chosen by the proposed algorithm. The green dashed line is value of $RepRatio_{gt}$.