Subject category:
Marketing
Published by:
Darden Business Publishing
Version: 23 August 2019
Length: 11 pages
Data source: Published sources
Topics:
Collaborative filtering; Slope one; Parametric model; Non-parametric model; Logistic regression; Weighted averages; Predictive analytics; Ordinal logit; Recommendation algorithms; Choice behavior; Decision analysis; Proportional odds assumption; Alternative least squares; Cold start problem; Popularity bias
Abstract
Netflix Top Picks, Amazon recommendations, the iTunes Genius button. They all have one thing in common: they are driven by clever algorithms that use a technique known as collaborative filtering. Often used in machine learning operations, collaborative filtering is the process by which a firm like Netflix generates predictions about a single user's preferences using data taken from a large number of users. This technical note offers an overview of three of the main collaborative filtering methods: slope one, a purely predictive nonparametric model; ordinal logit, a parametric regression model; and alternative least squares, a matrix factorization technique.
About
Abstract
Netflix Top Picks, Amazon recommendations, the iTunes Genius button. They all have one thing in common: they are driven by clever algorithms that use a technique known as collaborative filtering. Often used in machine learning operations, collaborative filtering is the process by which a firm like Netflix generates predictions about a single user's preferences using data taken from a large number of users. This technical note offers an overview of three of the main collaborative filtering methods: slope one, a purely predictive nonparametric model; ordinal logit, a parametric regression model; and alternative least squares, a matrix factorization technique.