Recommender system technology is the core of Netflix and Amazon’s business model and has lead to a tremendous increase in sales and customer satisfaction. Other retailers have seen sales increases of 5-15%, and now recommender systems are making their way to other industries to help customers find products faster, help salespeople find collateral and configure solutions, and help companies accelerate their product development by finding the right components to make products that meet market needs.
Real-time recommender systems are one of the sweetspot use cases for native graph databases. Key goals for a good recommender system include relevance, novelty, serendipity and recommendation differentiation. In this talk, Pieter will demonstrate how you can have full and accurate control of the recommender system with Neo4j, interactive response at scale, and “on the fly” tuning for a fast time to market.
– Why graphs are a really good fit for recommender systems
– How to ensure your recommender system is agile friendly
– How to incorporate Relevance, Serendipity, Novelty and Recommendation diversity
– White box recommender (opposite of black box)
– The power of hybrid recommender systems