| With the rapid development of the Internet,the data size of the online news is constantly increasing.Nowadays,the usage amount of mobile devices spread rapidly with mass popularization,which makes journalists write news easier,readers get information more convenient.In the process of the news production and consumption,a mass of data have been generated.Research on the data mining strategy of the data in network news has significant value both in the aspect of theory and application.Because of the huge amount of the Internet news,news producers may face lots of difficulties when mining the knowledge of news production system.Meanwhile,for the news consumers,it is much harder to find high value news conforming with their interests in a large number of Internet news.The former relies on efficient and accurate topic clustering algorithms of the Internet news,while the latter relies on excellent personalized recommendation algorithms.To this end,this paper proposes a new data mining and recommendation algorithm for the production system and the consumption system in the Internet news system,which includes:1)For the existing problems,such as semantics confusion and the difficulties in the incremental update of the clustering algorithm,this paper proposes Single-Pass clustering algorithm based on the LDA topic model.Through the rational use of news headlines,text and clue document,this paper realizes the incremental updating and improves the effect of news topic clustering algorithm by combining the improved LDA topics clustering strategy with Single-Pass clustering algorithm.Simulation results show that this algorithm improves the accuracy of clustering algorithm.And this algorithm is suitable for incremental updating.2)Aiming at the poor timeliness and redundancy of old news in the Internet news recommendation system,this paper combines the literature information aging model,proposing a recommendation algorithm based on the news topic clustering results.The paper calculates the aging rate of news after released with negative exponential model.Also,the news aging rate during collaborative filtering recommendation and the prior recommending high timeliness news have been taken into consideration,From the verification of the simulation results,improvement in the effect of collaborative filtering algorithm can be achieved with the new algorithm. |
