A consumer interest prediction system from transaction behaviors in electronic commerce

Consumer interest prediction usually uses transaction behaviors for predicting consumer's goal and interested items. The correct prediction heavily depends on the complete information of user profiles. The prediction system may conduct wrong prediction if it only uses the static, passive, incomplete, and out-of-date consumer transaction information to do the prediction. This paper proposes a consumer interest prediction system from transaction behaviors in electronic commerce. The system contains three modules, namely, transaction analyzer, transaction case library, and plan predictor. The transaction analyzer monitors the interactions of the consumer in the application systems. The transaction case library stores instances of consumer transaction behaviors in electronic commerce. The plan predictor integrates induction and decision theories for predicting consumer interest.