Estimating Probabilities in Recommendation Systems

TL;DR

This paper introduces a non-parametric kernel smoothing method for estimating probabilities in recommendation systems, interpreting missing data as censored observations, and demonstrates its effectiveness with real-world movie data.

Contribution

It presents a novel probabilistic estimation approach using kernel smoothing and censored data interpretation, enhancing recommendation systems with probabilistic preferences.

Findings

Comparable performance to state-of-the-art methods

Provides probabilistic estimates beyond traditional systems

Efficient computation schemes using generating functions

Abstract

Recommendation systems are emerging as an important business application with significant economic impact. Currently popular systems include Amazon's book recommendations, Netflix's movie recommendations, and Pandora's music recommendations. In this paper we address the problem of estimating probabilities associated with recommendation system data using non-parametric kernel smoothing. In our estimation we interpret missing items as randomly censored observations and obtain efficient computation schemes using combinatorial properties of generating functions. We demonstrate our approach with several case studies involving real world movie recommendation data. The results are comparable with state-of-the-art techniques while also providing probabilistic preference estimates outside the scope of traditional recommender systems.