A heuristic as basis for an adaptive e-commerce recommender system

Abstract

The research described in this paper proposes an evolution heuristic for realising an adaptive semantic e-commerce recommender system by establishing a feedback cycle. This recommender extracts questions from product domain ontologies (PDO) which are used in the dialogue of the recommendation process. The heuristic decides an automated PDO evolution (without a human inspection) in order to realise an automatic adaptation of the recommendation process. The feedback is derived from user interactions with the user interface of the recommender. This research shows that the automated PDO evolution outperforms a manual one. The evolution heuristic has been evaluated with an experiment and validated in real-world testing series.