What is it about?
This work is about measuring the change in the performance of a recommendation system resulting due to the change in the dataset properties. Primarily, the reduction in the performance of a recommender system due to subsampling compared to that of original data is formulated based on the properties of the original data set and subsampled fraction of users and items.
Featured Image
Photo by Muhammad Rizki on Unsplash
Why is it important?
Subsampling techniques help to reduce the computational cost of a recommendation technique but affect its performance. So it is essential to measure the effect of subsampling on the performance of a recommender system to optimize the levels of subsampling.
Perspectives
The formulas derived in this work to measure the effect of subsampling on SVD recommendation method can be taken as base for developing theories for other recommendation system algorithms.
Samin Poudel
North Carolina Agricultural and Technical State University
Read the Original
This page is a summary of: Closed-Form Models of Accuracy Loss due to Subsampling in SVD Collaborative Filtering, Big Data Mining and Analytics, March 2023, Tsinghua University Press,
DOI: 10.26599/bdma.2022.9020024.
You can read the full text:
Contributors
The following have contributed to this page
