Friday 15:00–16:30 in Track 2

Search Relevance: A/B testing to Reinforcement Learning

Arnab Dutta

Audience level:
Novice

Description

In order to achieve optimal ranking algorithm in an e-commerce platform, a single algorithm might fall short in order to achieve the desired results. We compare and contrast, A/B testing and reinforcement learning based techniques in our search for optimal algorithm.

This is the Tutorial Repo Link

Abstract

Ranking is critical in any e-commerce platform. An optimal ranking should not only be user relevant but also target at maximizing the business KPIs. In order to achieve that, often a single optimal ranking algorithm might fall short in order to achieve the desired results. Usually optimal algorithm selection takes place with A/B testing framework, however using reinforcement learning techniques, can often achieve similar and even better results under certain circumstances. In this tutorial, we go through the basics of the standard methodologies for finding a better ranking strategy, and implement from scratch a reinforcement learning framework to find an optimal ranking method. In the process we compare and contrast the techniques and outline the scope of these methodologies.

Subscribe to Receive PyData Updates

Subscribe