| As we all know,the development speed of the Internet in recent years beyond people’s imagination,the user left countless information on the internet.How the Internet information the excess in the dug out the content of interest to the user,and take the initiative to recommend to the user,and this is the main task to solve the problem of recommendation system.With the increasing number of users,a variety of portals,e-commerce sites and major companies of the system to provide users with more and better information services.E-commerce sites,for example,in the face of a variety of online shopping information,for the user to know what they want,you can take advantage of e-commerce site search engines to find their own products to buy.However,the user’s needs are often uncertain and vague,because sometimes they also do not describe what they want.At this time,if the shopping site can accurately recommend those users clear description of commodity information,can cultivate consumers rely on their own website,bring huge profits for their own website,so as to gain a firm foothold in the fierce competition in the commercial market.How to find the user’s needs in large amounts of information and recommend to the customer,is a major challenge facing the Internet companies,many scholars and experts to become a hot research topic.In the context of this business driven and academic atmosphere,the application of recommendation systems based on large data.Recommendation system can solve the problem of "information overload" in today’s Internet,which is popular among major companies and e-commerce websites.However,the recommendation system is not a panacea,in the face of more and more different types of data in the Internet and more and more complex application scenarios,the recommendation system is also facing many problems to be solved.These are some of the problems that have not been solved well in theory and application.The focus of the research in the field of recommendation system is focused on the scalability of the system,the cold start problem of new users and newprojects,data sparseness and so on.Data sparseness problem has become a bottleneck in the development of recommender systems.The existence of this problem seriously affects the recommendation quality of recommender systems.How to solve the problem of data sparseness becomes the key to guarantee the quality of recommender systems.Recommended system data sparsity problem arises because in the recommendation process,users need to rely on the project’s score data mining user interest information,to make recommendations to the user,the data dependence affects the accuracy of recommendation.The more data you rely on,the more accurate the results you recommend.However,the reality is contrary to the past,and users of recommender systems often leave no scoring data for some reasons,and the amount of data that the recommender system can rely on is scarce.Therefore,it is difficult for the recommendation system to find similar users according to the data,which leads to the low accuracy of the recommendation system and makes the target users dissatisfied.This thesis is devoted to the research of data sparsity in recommender systems.Based on the previous research,the data sparseness problem is studied.The main research work of this paper is as follows:This paper expounds the reasons for the data sparseness of recommender system,and analyzes its influence on recommender system recommendation accuracy.The shortcomings of existing methods to solve the problem of data sparseness are pointed out.Fixed filling method does not consider the attributes of users and projects,will bring the deviation on the recommendation accuracy;matrix reduction method is easy to lose data,makes predictions of the score is not accurate;Content-based CF has no data sparseness problem,but can not find new users of potential interest,recommendation system is meaningless.By analyzing the respective advantages of User-based,CF and Item-based CF,this paper proposes an idea that combines the traditional CF fill matrix to solve the problem of data sparsity in recommender systems.The idea not only alleviates the data sparsity problem of recommender systems,but also improves therecommendation performance and improves the recommendation quality of recommender systems.Apache Mahout recommendation technology framework and Movie Lens data set for experiment,and the traditional collaborative filtering recommendation algorithm are analyzed and compared,proves that the combination of new traditional CF filled matrix idea can alleviate the sparsity problem,greatly improve the recommendation quality of recommendation system. |
