DS Evidence Theory And Shannon Entropy Social Tag Aggregation Recommendation Algorithm

At present, the recommended search research focused on tag recommendation in social tagging system. That is a given a user and a resource, recommendation system to predict user with what tags to explain the resources, how to develop resource oriented personalized recommendation system has been filled with the need of application, however, there are still a lot of problems. There have been many mature recommendation system, user-based CF, graph-based methods, social-based CF, although these system has been applied to large degree, but these have some disadvantages, makes recommendation effect defective. In order to recommend better, first we can think of is the mixed strategy known to produce a hybrid recommendation algorithm. Mixed strategy recommendation algorithm is divided into three kinds: integration strategy, parallel strategy, pipe strategy. We use the parallel strategy, as the name implies, is the recommended list are generated recommendation algorithm for each individual, then the algorithm fusion of these recommended list forming recommended list final polymerization through specific data.Our main work is the aggregation algorithm, there are a lot of aggregation algorithms, such as BordaCount, CombSUM, we introduce a novel aggregation algorithm using DS evidence theory and the Shannon entropy grade. But in aggregate the different sources of information, there will be information conflict problems, in order to solve the information conflict problem, we use the DS evidence theory and the Shannon entropy to the polymerization of these class list, so you can achieve a more accurate and reliable rank list. We implemented our algorithm in Lastfm data sets, the result shows, our method can achieve a more accurate performance, confirmed the ability of the algorithm we apply the framework.