Logistic Matrix Factorization for Implicit Feedback Data

Abstract

Collaborative filtering with implicit feedback data involves recommender system techniques for analyzing relationships betweens users and items using implicit signals such as click through data or music streaming play counts to provide users with personalized recommendations. This is in contrast to collaborative filtering with explicit feedback data which aims to model these relationships using explicit signals such as user-item ratings. Since most data on the web comes in the form of implicit feedback data there is an increasing demand for collaborative filtering methods that are designed for the implicit case. In this paper we present Logistic Matrix Factorization (Logistic MF), a new probabilistic model for matrix factorization with implicit feedback. The model is simple to implement, highly parallelizable, and has the added benefit that it can model the probability that a user will prefer a specific item. Additionally, we show it to experimentally outperform the widely adopted Implicit Matrix Factorization method using a dataset composed of music listening behavior from streaming music company Spotify

Related

July 2021 | SIGIR

Podcast Metadata and Content: Episode Relevance and Attractiveness in Ad Hoc Search

Ben Carterette, Rosie Jones, Gareth Jones, Maria Eskevich, Sravana Reddy, Ann Clifton, Yongze Yu, Jussi Karlgren and Ian Soboroff

July 2021 | SIGIR

Current Challenges and Future Directions in Podcast Information Access

Rosie Jones, Hamed Zamani, Markus Schedl, Ching-Wei Chen, Sravana Reddy, Ann Clifton, Jussi Karlgren, Helia Hashemi, Aasish Pappu, Zahra Nazari, LongQi Yang, Oguz Semerci, Hugues Bouchard, Ben Carterette

May 2021 | CHI

Towards Fairness in Practice: A Practitioner-Oriented Rubric for Evaluating Fair ML Toolkits

Brianna Richardson, Jean Garcia-Gathright, Samuel F. Way, Jennifer Thom, Henriette Cramer