WEB recommendations for E-commerce websites

In this part of the thesis we have investigated how the navigation utilizing web recommendations can be implemented on the e-commerce websites based on integrated data sources. The integrated e-commerce websites are an interesting use case for web recommendations. One of the reasons for this interest is that many modern, large and economically successful e-commerce websites follow the integrated approach. Another reason is that especially in the integrated environment, due to the lack of the pre-defined semantic connections between the data, the web recommendations step forward as means of enabling user navigation. In this chapter we have presented the architecture for the websites based on integrated data sources named EC-Fuice. We have also presented the prototypical implementation of our architecture which serves as a proof-of-concept and investigated the challenges of creating navigation on an integrated website. The following issues were addressed in this part of the thesis: Combination of several state-of-the-art tools and techniques in the fields of databases, data integration, ontology matching and web engineering into one generic architecture for creating integrated websites. Comparative experiments with several techniques for instance matching (also known as record linkage or duplicate detection). Investigation on using the ontology matching to facilitate the instance matching. Comparative experiments with several techniques for ontology matching. Investigations on the instance-based ontology matching and the possibilities for combining instance-based ontology matching with other techniques for ontology matching. Investigation of the possibilities to improve user navigation in the integrated data environment with different types of web recommendations. Review of the related work in the fields of data integration and ontology matching and discussion of the contact points between the research described here and other related projects. The main contributions of the research described in this part of the thesis are the EC-Fuice architecture, the novel method for matching e-commerce ontologies based on combination of instance information and metadata information, the experimental results of ontology and instance matching performed by different matching algorithms and the classification of the types of recommendations which can be used on an integrated e-commerce website