Research On Personal Recommendation System Based On Tags

In recent years,with the vigorous development of computer and network technology,information overload of Internet has become increasingly serious,which lead to a massive information space where people can't obtain the information they need efficiently.At present,as the most effective tool to solve problem of information overload and help users acquire information,recommendation system has gained a rapid development and achieve great success in many fields.At the same time,with the advent of the of web2.0 era,the recommendation algorithm based on label has been paid wide attention by researchers gradually.Traditional recommendation algorithm,however,there are always problems of data sparse and cold start,which could lead to an inaccurate recommendation result.In order to deal with the problem of cold start and improve recommendation accuracy,Based on word2 Vec technology,this paper makes word2 Vec experiment,training word vector to calculate the text similarity.Then,by the text similarity the paper get a hybrid recommendation algorithm considering aspects of user behavior and item content,which can improve the existing recommendation algorithm,because the text similarity can present the content relevance between items.At first,the paper comprehensively introduces the recommendation system theory including the various modules,evaluation method and applications of recommendation system.Then,the paper expounds the concept of tag and the significance of tag to recommendation system,on this basis the paper introduces the commonly used tag recommendation algorithms.Later,the paper introduces related principles of word2 Vec technology.As an effective tool of training word vector that represent text semantic,word2 Vec can help us understand and deal with many nature language processing problems.On the basis of training word vector with the tool of word2 Vec,the paper proposes the concept of text similarity,by which we can present the content relevance of items.Then the paper gets a improved algorithm based on the text relevance,which can ameliorate the items start problems of recommendation system on the one hand and improve the accuracy of recommendation result by filtering the items which has low relevance.At last,thepaper makes simulation experiment,verities the validity of the improved algorithm from multiple evaluation indicators.