MFID: An Improved Matrix Factorization Model for Incremental Data

Matrix factorization, a widely used technique in various data analysis domains, is extended to address the challenges of incremental data updates while maintaining computational efficiency. Our novel approach, termed Matrix Factorization for Incremental Data (MFID), focuses on updating a single factor matrix, either rows or columns, in response to the incremental addition of data. This innovative adaptation significantly reduces computational overhead compared to traditional methods that require full re-factorization of the matrix. The MFID model combines principles of online learning with advanced update strategies to effectively capture evolving patterns in dynamic datasets, making it a valuable tool for applications such as real-time recommendation systems, collaborative filtering, and dynamic image analysis. Extensive experimentation demonstrates its superior performance in scenarios demanding both row and column updates, highlighting its practical applicability in various domains.