Recency Dropout for Recurrent Recommender Systems

TL;DR

This paper introduces recency dropout, a data augmentation method designed to reduce recency bias in recurrent recommender systems, improving their ability to model long-term user interests.

Contribution

The paper proposes recency dropout, a novel technique that mitigates recency bias in RNN-based recommenders, validated through simulations, offline, and live industrial experiments.

Findings

Recency dropout improves long-term user interest modeling.

It reduces recency bias effectively in various settings.

Enhances overall recommendation quality and user experience.

Abstract

Recurrent recommender systems have been successful in capturing the temporal dynamics in users' activity trajectories. However, recurrent neural networks (RNNs) are known to have difficulty learning long-term dependencies. As a consequence, RNN-based recommender systems tend to overly focus on short-term user interests. This is referred to as the recency bias, which could negatively affect the long-term user experience as well as the health of the ecosystem. In this paper, we introduce the recency dropout technique, a simple yet effective data augmentation technique to alleviate the recency bias in recurrent recommender systems. We demonstrate the effectiveness of recency dropout in various experimental settings including a simulation study, offline experiments, as well as live experiments on a large-scale industrial recommendation platform.