Presented By O'Reilly and Cloudera
December 5-6, 2016: Training
December 6–8, 2016: Tutorials & Conference
Singapore

Context-aware recommendations using reinforcement learning in the item-similarity space

Arun Veettil (Skellam AI)
2:35pm–3:15pm Wednesday, December 7, 2016
Chat, machine learning, and AI
Location: Summit 1 Level: Intermediate
Average rating: **...
(2.20, 5 ratings)

Prerequisite Knowledge

  • Basic knowledge of big data echo systems, such as Spark and Hadoop
  • A basic to intermediate understanding of statistics and probability distributions

What you'll learn

  • Understand how to build context-aware recommendation engines with reinforcement learning
  • Explore a big data echo system for building such a recommendation engine
  • Learn how to perform dynamic optimization for improving conversion rate

Description

Unlike traditional recommendation algorithms, which use explicit and implicit user ratings and apply some sort of collaborative filtering techniques, product recommendations in the food and beverage industry have to be sensitive to the context of the recommendation (location, time, day, etc.). In addition, recommendations have to be very accurate in terms of ranking, as customers are hardly interested in browsing through a web of products (unlike in movies and apparels) when they expect the algorithm to accurately identify patterns in their food habits and respond accordingly.

The problem is how to incorporate the contexts of tens of millions of customer base and then incorporate reinforcement learning to track the behavioral changes in customers preferences. Arun Veettil explains how to incorporate user contextual information into recommendation algorithms and apply reinforcement learning to track continuously changing user behavior.

Topics include:

  • The algorithm behind product recommendations using context-specific reinforcement learning
  • How big data echo systems like Spark and Hadoop are helping to power these algorithms
  • How feedback loops are incorporated into recommendations
  • How to dynamically optimize for revenue or click-through rate for each customer
  • Lessons learned and pitfalls encountered
Photo of Arun Veettil

Arun Veettil

Skellam AI

Arun Veettil is the chief architect at Skellam AI, a company dedicated to helping businesses develop custom AI solutions. Arun is a tech veteran with over 17 years of industry experience. For the last seven years, Arun has been working at the intersection of machine learning and product development. Previously, he worked at Point Inside, Nordstrom Advanced Analytics, the Walt Disney Company, and IBM. His expertise includes developing machine learning algorithms to run against very large amounts of data and building large-scale distributed applications. Arun holds a master’s degree in computer science from the University of Washington and a bachelor’s degree in electronics engineering from the National Institute of Technology, Allahabad, India.