Weight-Vector Based Approach for Product Recommendation in E-commerce

Abstract

This paper presents a knowledge-based product retrieval and recommendation system for e-commerce. The system is based on the observation that, in Business to Customer (B2C) e-commerce, customers’ preferences naturally cluster into groups. Customers belonging to the same cluster have very similar preferences for product selection. The system is primarily based on product classification hierarchy. The hierarchy contains weight vectors. The system learns from experience. The learning is in the form of weight refinement based on customer selections. The learning resembles radioactive decay in some situations. Labor profile domain has been taken up for system implementation. The results are at the preliminary stage, and the system is not yet evaluated completely.