Exploiting knowledge about fashion to provide personalised clothing recommendations

In the fashion industry, mass-customisation is a new trend that tries to produce clothes respecting the idiosyncracy of every customer and doing so cost effectively. In this paper we present a knowledge framework that leverages the above process by providing personalised clothing recommendations. The methodology that we propose, and the prototype we have built, incorporates knowledge about aspects of fashion, such as materials, garments, colours, body types etc. into a ontology. With the aid of concepts and relations of the ontology, domain experts can also define style advice rules. Moreover, a general- purpose personalisation server (PServer) is employed, that stores style advice rules in the form of user stereotypes and mines user interaction data to produce patterns that enrich the experts' style advice rules. Due to the synergy of the domain ontology and the PServer there is an impetus to map style advice rules between the two different representations. Finally, a recommendation engine that exploits user stereotypes is built in order to suggest new fashion items to users.