Journal des technologies de l'information et du génie logiciel
Libre accès

Abstrait

Proposal of a Hybrid Recommendation Algorithm to Support the Discovery for Mashup Applications

Takahiro Koita, Daiki Takigawa

This paper proposes a recommendation algorithm which combines collaborative filtering and content based algorithm. The proposed algorithm provides recommendation list that combines recommendation items generated each algorithms, and improves the novelty and the precision of recommendation. Especially, if the precision is low, the content-based algorithm should have higher priority and if the precision is high, the collaborative filtering should have higher priority. Therefore, this paper discusses and investigates priority rules and priority through the preliminary experiments. The priority rules are some rules to decide priority algorithm when combine two existing algorithms. The priority is a weight for priority algorithm. To decide appropriate priority rules and priority, the proposed algorithm was implemented on the Linked Mash which is our recommendation system of mashup applications and we conducted experiments with Linked Mash. In the experiments, the subjects evaluated some recommended mashup applications. The novelty and precision is calculated based on this evaluation. Changing the priority rules and priority for each subjects, we demonstrated that proposed algorithm can achieve a recommendation which both novelty and precision are high