cohort

Podcast recommendations from listening-app patterns.

Co-clustering applied to public podcast download data. Groups shows and listening apps simultaneously, surfacing audience patterns that single-axis methods miss.

Analysis
The cohort story. Data, findings, and how the recommender stays current.
Algorithm
Math, predictor, training loop, notation, and our adaptation.
Explore the paper ↗
George & Merugu, A Scalable Collaborative Filtering Framework Based on Co-Clustering, ICDM 2005.