Online Learning for Recommendations at Grubhub

TL;DR

This paper introduces an online learning method for recommender systems at Grubhub, leveraging transfer learning to improve adaptability, reduce costs, and enhance recommendation quality in a rapidly growing platform.

Contribution

It presents a practical approach to convert offline recommenders to online, addressing challenges like convergence and non-stationary data, with demonstrated cost savings and performance improvements.

Findings

Up to 45x cost savings in platform operations

Over 20% increase in recommendation metrics

Effective handling of concept drift in production systems

Abstract

We propose a method to easily modify existing offline Recommender Systems to run online using Transfer Learning. Online Learning for Recommender Systems has two main advantages: quality and scale. Like many Machine Learning algorithms in production if not regularly retrained will suffer from Concept Drift. A policy that is updated frequently online can adapt to drift faster than a batch system. This is especially true for user-interaction systems like recommenders where the underlying distribution can shift drastically to follow user behaviour. As a platform grows rapidly like Grubhub, the cost of running batch training jobs becomes material. A shift from stateless batch learning offline to stateful incremental learning online can recover, for example, at Grubhub, up to a 45x cost savings and a +20% metrics increase. There are a few challenges to overcome with the transition to online stateful learning, namely convergence, non-stationary embeddings and off-policy evaluation, which we explore from our experiences running this system in production.