Thesis Public Defense | VNP27 - Trần Ngọc Sơn

Price-aware matrix factorization for recommend system: A case study in Vietnam

Student: Trần Ngọc Sơn, VNP-27

Supervisor: Dr. Trần Thị Tuấn Anh

Abstract:

In recent years, Vietnam has witnessed a significant transformation in consumer behavior due to the rapid expansion of the Internet and the far-reaching effects of the COVID-19 pandemic. Due to the convenience of e-commerce platforms, the growing availability of Internet services, and affordable prices, online shopping has become necessary for daily life. This shift has propelled the e-commerce industry’s continuous growth, turning it into a virtual marketplace with a diverse range of products. Advanced recommendation systems facilitate this evolution, enhancing personalization and the shopping experience. This study underscores the critical role of pricing in shaping consumer decisions, particularly during economic challenges when a significant portion of customers prefer budget-friendly retail options. The study uses the price-aware Singular Value Decomposition (PASVD) model, which considers people’s preferences and price sensitivity when making suggestions. The primary research question is, “How can the PASVD model be effectively optimized and applied to new datasets within Vietnam’s e-commerce context?” This investigation focuses on parameter optimization to improve accuracy and efficiency in tasks related to rating prediction, aiming to enhance the customer experience in Vietnam’s burgeoning e-commerce sector. This study shows that the PASVD model helps accurately suggest products in the Vietnamese e-commerce industry.

Keywords: E-commerce, Price Preference, Price Sensitivity, Personalized Recommendation, Recommender System, Collaborative Filtering, Matrix Factorization, Data Sparsity, The Number of Latent Factors, Regularization Coefficient.