Abstract:
Personalized service is becoming a key strategy in electronic
commerce. Traditional personalization techniques such as
collaborative filtering and rule-based method have many
drawbacks, including lack of scalability, reliance on subjective
user rating or static profiles, and the inability to capture a
richer set of semantic relationships among objects. In this
paper, we present a new approach, building customer profiles
based on product hierarchy for more effective personalization
in electronic commerce. We divide each customer profile into
three parts: basic profile, preference profile, and rule profile.
Based on the customer profiles, two kinds of recommendations
can be generated, which are interest recommendation and
association recommendation. We also propose a special data
structure: Profile Tree for effective searching and matching.
In terms of our method, customer profiles can be constructed
online, and realtime recommendations can be implemented. In
the end, we conduct experiments to validate our methods,
using real data.
