Research On Demand - Oriented User Modeling And Service Recommendation

The technique of personalized service recommendation has been widely used. However, most user models heavily depend on the data from the Internet such as user’s rating and click behaviors. User models with low data dimension or no dimension separation is incomplete and low coherent, which brings the difficulty of accurately describing user’s demands and preference.Based on the project requirements of an e-tourism platform and the research of different kinds of user model, we propose a user model that is based on the multi-dimensional description of user’s demands. The goal of the model is to describe the user with deep completeness by exploiting the blooming data from sources such as wearable devices and the net of thing. The recommendation can support goal-oriented reasoning, and thus has a better performance.The proposed model can fully reflect the user demands by describing the user from 6 aspects, e.g. physics, personality traits, attitudes, knowledge, experience and environment. Based on the model, a novel recommendation method based on user’s demands is proposed which exploits BDI agent reasoning framework. The belief of the agent includes the user model and domain knowledge. The desire of the agent is used to represent the user demands and the satisfaction conditions of these demands. The method simplifies the semantic representation by using multi-dimensions, and the user demands are derived based on the cross-validate between dimensions. Services are recommended to the user on how they satisfy the user’s demands.A prototype of travel service recommendation system is implemented to practice the proposed model and recommendation method. The implementation is based on an open source BDI platform named Jadex. The prototype system is deployed on the smart e-tourism portal and is proved to have retrieved periodic progress.