Inside the Co-Occurrence Recommendation Engine

Join us as we discuss recommender systems! Soft drinks and light snacks will be provided.

----------------------------------------------------------------------------------------------------

What is a recommender system?

Online shopping provides the means for a business to present a vast number of products to consumers. In order to increase sales, it is desirable to present a focused list of “recommended” products to a user that the user would be interested in purchasing. To create a list of recommended products, the historical purchase history and user demographics need to be processed.

This presentation provides an in-depth analysis of the inner-workings of a recommender called “co-occurrence”. This type of recommender is simple, yet it is powerful enough to be used for various applications. This presentation leads by way of example to show the various steps used to create a recommendation system. Once the co-occurrence matrix is computed, two different styles of system can be developed: (1) an individual recommender that takes the current user items and creates a recommendation for the current products, and (2) a search engine recommender that finds surprising relationships between items and presents them as additional items in a search response. The co-occurrence recommender can provide cross-recommendation (i.e. browsing movies can recommend music) and can be extended for streaming recommendations (matrix is updated as each new item arrives).

Spend a little time coming up to speed on the core technologies of a recommender system and think about how to reach your customers!

----------------------------------------------------------------------------------------------------

Speaker Bio:

Brett Lindsley is currently a Solutions Architect III at Grainger. At Grainger, he develops e-commerce systems for the MRO industry. His previous position was a Senior Manager at Gogo developing aviation software. The biggest system at Gogo was a big data system for flight data that merged millions of data sensor observations into flight records every day. Previously Brett worked at BuildingWorx on smart building software as an Architect and Data Scientist. As a Data Scientist, Brett developed new algorithms for the modeling and prediction of energy consumption based on time/temperature modeling. Brett’s previous position was with Motorola in the Applied Research Center as a Distinguished Member of the Technical Staff. At Motorola, Brett developed advanced products and technology for 28 years. During this time, Brett has produced 24 issued patents and 4 publications in technologies ranging from digital signal processing, IC design, video systems and enterprise software. Brett has six certifications – AWS Certified Solutions Architect Associate, FAA Fundamentals Of Instruction (FOI), SCJP, SCWCD, SCBCD and SCMA. Brett enjoys system and application development with Java and Spring and presenting at the Java User Group (JUG). Brett has BS/MS degrees from North Dakota State University in Electrical/Computer Engineering.