Apache Spark 2.x Machine Learning Cookbook

Content filtering

Content filtering is one of the original techniques for recommendation engines. It relies on user profiles to make recommendations. This approach relies mostly on pre-existing profiles for users (type, demographics, income, geo-location, ZIP code) and inventory (characteristics of a product, movie, or a song) to infer attribution which then can be filtered and acted upon. The main issue is that the pre-existing knowledge is often incomplete and expensive to source. This technique is more than a decade old and is still being practiced.