Learning the users' preferences in e-commerce: A weight-adjustment approach

摘要

Learning the preferences of users is an important problem in e-commerce research. This paper presents a system for that purpose, and it is primarily based on weight vectors. Learning is incorporated in the form of refining the weights. The system is sensitive to the users' change of trend and is implemented for labor profile domain. An empirical evaluation has been conducted in a simulated environment. The results proved the following: (1) The system converges if the user populations have some common preferences, (2) The system detects and adapts to a change of trend, and (3) It takes more time to converge in the case of more weights.