Research For Ontology And Knowledge Rule Based Hybrid Recommendation

The increasing development of network technology has provided massive information on the web, which on the contrary become a burden for the internet users to search what they want, mostly because the current network applications has not taking the individual needs of users into account. Although the search engine allows users to use keyword to query items of their interest, the service is passive, and the results returned may not be satisfactory. The appearance of personalized recommendation system effectively solved this difficulty, it can not only find the user’s interest automatically, but also recommend projects of interest to the users, thus become a new approach to internet service.In general, in order to make a recommendation system, a designer can consider using two approaches:content-based filtering approach and collaborative filtering approach. However, they both have some their own technical shortcomings. The content-based approach is difficult to handle feature extraction and user intension prediction. The collaborative approach faces the cold start problem and the matrix sparsity problem. In this paper, we combines content-based filtering, collaborative filtering and the new tool ontology to present an novel hybrid recommendation approach. The new recommendation approach can solve the traditional recommenders’ problems, such as feature extraction, intension prediction, matrix sparsity and cold start problems.All work of this paper is to solve the shortage of the traditional recommendation algorithm and improve the quality of the recommendation of recommendation system. In this paper we use the new tool-ontology to improve the traditional recommendation algorithm. First, we put the ontology to construct the movie domain’s user model and Item model. It makes the keywords that express the user interest and the Item content in a same domain ontology and makes all kinds of recommendation algorithm can use the same data model. Then, we build a preference estimating function. It can estimate the preference user to the video by analyzing of user behavior data. This function can estimate users ratings and reduce the matrix sparsity of the traditional collaborative filtering algorithm on one hand, on the other hand, can be used to automatically update the service for the user interest model. In This paper we also formulate a set of quick and accurate user interest update method, the new update method can find out the users new interests quickly and accurately. In addition, we also built the similarity calculation formula based on ontology in the field of film; it can make system cover the semantic relations when calculating the similarity between the keywords. Based on these, this paper presents a variety of recommendation algorithm based on ontology. In user interest update algorithm and recommendation algorithm this paper also takes into account the time factor, so that the system can quickly express the new interest to the user in the recommend results. Finally, we also introduced the knowledge rules to our system, which can make a recommendation system more complete and intelligent and can also make the recommended results by the system more humanized.