The Design And Implementation Of News Recommendation System Based On Hybrid Strategy

Network news has become the majority of Internet users most concerned resources,network news compared to the traditional newspaper and television news have advantages,such as have better real-time,easy to read at any time,without restrictions of time,but it is precisely these advantages make the network news unrestricted massive growth,the reader has to spend a lot of time to find their favorite news.Recommended system is to solve the problem of information overload,which makes the user does not have to spend a lot of time on information search,but automatically display information for the user.News recommendation system is one aspect of the recommendation system applications,It can show user interest news first.Enhance web services personalized and intelligent,also can increase user's loyalty for the site,but also saves time for the user.The main work includes two aspects,one is study recommendation algorithms the other is the recommendation system's design and implementation,Details are as follows:(1)Words semantic similarity calculation method applied to News recommendation system.Proposed Double Model method to Calculate semantic similarity between news and user.Double models include a wide range of user interest model and centralized user interest model.The parameter of the double models is the number of news and the number of feature words.the widely user interest model used large number of news articles and fewer number of features,centralized user interest model using fewer number of articles and large number of feature words.Calculate similarity between double model and news content model,and then make recommendations based on their similarity values.In experiments on real data sets,improved algorithm has a higher recommendation accuracy rate than the unmodified algorithm,and observe the effect of different values of model parameters on the accuracy of the recommendation.(2)Learm the principle of LDA model and its application in the news recommendation system.Establishment user interest model based on LDA model,proposed comprehensive user interest model based on widely interest and focus interest,and calculate similarity between comprehensive user interest model and news topic model,recommend high similar news to the user,experiments on real datasets found improved algorithm has a higher recommendation accuracy than the unmodified algorithm.(3)Bipartite graph theory applied in the field of news recommendation system.In the past,bipartite graph based on material and energy diffusion,have same initial energy.We consider the news popularity,try to given different initial energy value to different popularity news.In order to obtain the affect relationship between the news and the initial energy.Experiments on real data sets,and found popular news with a smaller energy value will bring higher recommendation accuracy rate.(4)Study and achieve the hybrid recommendation algorithm,Study and achieve the hybrid recommendation algorithm,which contains five preference factor,user preference for the first three algorithms,news popular preferences factor,time preference factor,the system combine the first three results and news popularity,taking into account the user real?time requirements of the news,to produce the final result.When the user read news,the first three preference factors will adjust,the proportion of the three algorithms in the recommend list is different.Experimental results show that the hybrid algorithm has better recommendation results.(5)Designed and implemented a news recommendation system based on hybrid strategy.By analyzing the user's reading recorded automatically recommend the user interest news.System uses three algorithms to recommended:algorithm based on LDA model,algorithm based on Words semantic similarity,algorithm based on Bipartite graph.The final recommendation is the result of a comprehensive of this three algorithms.The system includes the following modules:a data storage module,pretreatment module,a recommendation algorithm module,recommendation result integrated module,the results module.System uses large service software's real datas for testing and achieved good recommendation accuracy.