Research On Resource Recommendation Using Hybrid Recommendation Algorithm Based On Social Tags

The development of the Internet is so rapid recently,which has led to more and more serious information overload.It is becoming more and more difficult for people to find the information they are interested in.In order to solve this problem,an effective way is to use personalized recommendation to recommend resources for different users.The social tags can reflect the user's interest,as well as some of the information hidden in the resources.The introduction of social tags for personalized recommendation will lead to better results.The current research on personalized recommendation has reached remarkable achievements,however,there are still some disadvantages,for example,recommendation algorithms based on content are often only limited to the content available.If automatic feature extraction from multimedia data(images,audio,videos)is needed,it would be hard to do so due to the lack of sufficient text content.On the other hand,the collaborative filtering recommendation algorithm has a problem of data sparseness since the number of users who grade the resources is far less than the total number of all users.In order to solve these problems,we can introduce social tags.When it comes to the problems of semantic ambiguity and redundancy existed in social tags,we studied how to filter the noise information through group wisdom and proved that selecting appropriate effective tags is effective.In addition,we could make some changes to the TF-IDF algorithm to calculate the weight of social tags,and use the Dijkstra algorithm to calculate the distance between two resources,making improvement to the problem of data sparseness because possible relationship through the link between two seemingly unrelated resources may be found.At last,we construct a hybrid recommendation algorithm based on social tags by combining recommendation methods based on content and collaborative filtering algorithm.The hybrid recommendation algorithm we constructed is a combination of the two algorithms mentioned above,and it proves to have better result than each single recommendation algorithm.Based on the background of social tags and personalized recommendation technology,this paper studied the hybrid recommendation algorithm,and mainly accomplished the following work.First of all,this paper summarized the current research and development of social tags and personalized recommendation,and summarizes the related technologies and theoretical basis of resource recommendation algorithm.Secondly,this paper makes an in-depth analysis of the existing recommender systems and recommendation methods based on social tags,as well as the advantages and disadvantages of various methods and current unresolved research issues.Then,in this paper the existing resource recommendation method,collaborative filtering algorithm was improved by effective noise reduction of the traditional label is calculated by using the AF and CRAF values,and combined with the traditional Dijkstra algorithm recommendation algorithm based on TF-IDF content was improved,and the integration of the two algorithms and proposes a new hybrid Resource Recommendation method,and the detailed design of the recommended procedure.Finally,we selected the Movie Lens 20M datasets,respectively set the experimental group and control group,and conducted empirical research on the idea we constructed in this paper,which is resource recommendation using hybrid recommendation algorithm based on social tags.The experimental results are evaluated to verify the effect of this recommendation method.