A Dense Embedding Alignment Model for Resolving E-Commerce Catalog Sparsity

International Journal of Computer Science and Technology Vol. 3, No. 8, August 2001 | pp. 57–64

DOI: 10.46882/2001/IJCST/000032

Research Article

Title: A Dense Embedding Alignment Model for Resolving E-Commerce Catalog Sparsity

Names of Authors: Michael Brown¹, Sun-Woo Kim²

Authors’ Affiliations: ¹School of Computer Science, University of Manchester, UK; ²Department of Computer Engineering, Seoul National University, South Korea.

Abstract:
Digital commerce platforms depend heavily on information filtering layers to connect active buyers with relevant product listings. Standard correlation algorithms often fail when encountering severe cold-start parameters or limited historical transaction databases. This paper presents a novel hybrid prediction architecture that integrates user metadata descriptors with explicit rating vectors through an interconnected matrix layer. The model uses dense matrix embeddings to convert sparse categorical variables into a shared lower-dimensional operational layer, identifying complex connections between buyer behavior and catalog features. The hybrid model was tested using a public benchmark dataset containing 10 million data points. The outcomes show an overall Root Mean Squared Error (RMSE) reduction of 6.8% compared to baseline collaborative models, along with a 12.4% precision improvement for newly added catalog entries. The architecture displays strong resilience against data scarcity, maintaining accurate matches even when user profiles contain fewer than five history entries. This research proves that combining user profiles with item attributes resolves systemic data sparsity limitations.

Keywords: Recommender Paradigms, Collaborative Filtering, Matrix Factorization, Data Sparsity, Information Extraction, Retrieval Engines.

Manuscript Timeline: Received: April 02, 2001; Revised: May 15, 2001; Accepted: June 08, 2001; Published: August 20, 2001.

Citation: Brown, M., & Kim, S. W. (2001). A Dense Embedding Alignment Model for Resolving E-Commerce Catalog Sparsity. International Journal of Computer Science and Technology, 3(8), 57–64.

Subscribe to read the full article: https://internationalscholarsjournals.org/subscribe-to-read