The Research On Personalized Recommendation System Based On Social Tags

With the development of the Internet, the emergence of tag technology has brought new opportunities to the personalized recommendation. The social tags connected users and items, described the characteristics of the items, embodied the user preference, and brought vital data sources for the personalized recommendation system. However, the existing personalized recommendation system based on the tags still has some problems, which limited the application of tags in the areas of recommendation. Mainly in the following three aspects:(1)Semantic fuzziness problem of tags.because of fuzzy semantics, the user preference was difficult to describe, which reduced the precision of recommendation system.(2)Efficiency problem of recommendation system. With increasing of user annotation and expanding of the scale of calculate, the efficiency of recommendation system was reduced.(3) Quality problem of recommendation system. The quality of the recommending items was neglected.In order to solve the above problem, taking intelligent recommendation technology as background, taking tag-based personalized recommendation system as the research object, taking elimination for the tags semantic fuzziness problems and improvement of the accuracy, efficiency and quality of tag-based recommended recommendation system as research purposes, this paper mainly completed the following job:(1) Discovery of semantic topic based on social annotation. This paper analysed the relation among users, items and tags, introduced latent semantic analysis probabilistic PLSA model and extended it, the latent semantic of user's annotation was gained by method of discovery of semantic topic, this method mapped the annotation to well-defined semantic topic, the semantic fuzziness problem of tags was eliminated.(2) User interest model based on the semantic topic. On the basis of research on the discovery of the semantic topic, using semantic information which contained in semantic topic, this paper organized feature tags of user interest, then constructed the three-level user interest model based on semantic topic, and brought forward the refresh strategy to capture user interest changes. This model described user interest characteristics, and improved the precision of personalized recommendation system based on this model.(3) Cooperative filter recommendation algorithm based on the user interest model and item rating. This paper researched cooperative filter recommendation algorithm based on the user interest model and item rating, mined user potential interest with the aid of the user interested model, then recommended the items to the user within the scope of the user interest, the computing scale was reduced, so the efficiency of recommendation systems was improved, the quality of the recommendation systems was guaranteed with item rating.(4) This paper designed and developed a sample tag-based personalized recommendation system for book, introduced structure and recommend process of the recommendation system, realized most core functions of the system framework, verified part of performance and implementation effect of system by experiment.