Research On Key Technologies Of Military News Personalization Recommendation

In the age of informationization and big data,massive data has brought users the problem of data selection.In order to quickly feedback the data of interest to users,personalized recommendation technology emerges as the times require.Personalized recommendation technology is a good solution in information overload field.With the rapid development of sensors,the battlefield environment data is also increasing and showing an explosive growth trend.The growth of battlefield environmental data makes it difficult for traditional battlefield environmental security model to meet the needs of military personnel,and its security model gradually shifts from data-intensive to knowledge-intensive.In order to meet the challenges of the big data trend in the battlefield environment,provide knowledge services for military personnel,and improve the active protection level of the battlefield environment,this paper introduces personalized recommendation technology into the battlefield environmental data security,and discusses the key technical issues of personalized recommendation of military news.This paper will focus on three aspects: “recommendation model construction of military news”,“update of military news recommendation model” and “recommendation model of military news integrating geographic context”.The main works are as follows:1.The background of the battlefield environment data to provide security for military personnel was introduced,the problems existing in the current use of the “qipan” system by military personnel was analyzed.Comparing the data in the “qipan” system with the characteristics of the personalized recommendation objects,we found the geographical names data and Military news were applicable to the implementation of recommendations using personalized recommendation techniques.Therefore,military news was chosed as the research object of this paper,and the research significance of personalized recommendation of military news was summarized.The research status at home and abroad was analyzed,and the existing problems were summarized.Finally,the research ideas and organizational structure of this paper are summarized.2.In the aspect of user interest model construction,a news recommendation method based on Vector Space Model(VSM)and Bisecting K-means clustering was proposed.Firstly,the news feature vector was constructed by using VSM and TF-IDF algorithm.Then,the news feature vector set was clustered by Bisecting K-means clustering algorithm.Finally,the User-Categories-News User Interest Model(UCN-UIM)which is three-level hierarchical structure was constructed.This method not only preserves the interpretability advantages of the content-based recommendation algorithm,but also uses the clustering algorithm to avoid the problem of editing classification.3.The user interest model was updated with regard to the time context,and a time-based forgetting function was constructed.On the basis of work(2),the forgetting function was constructed based on the Ebbinghaus forgetting curve,and the UCN-UIM was time-weighted.The experimental results show that the updated model is better than the original model.4.A News recommendation algorithm Considering Geographical Position(NCGP)was proposed.Firstly,in order to extract the news event,a news event extraction algorithm was designed.Secondly,the method of constructing the news feature vector in work(2)was improved,on the basis of the original method,the word2 vec tool was used to construct the news feature vector.Then a User-Location-News User Interest Model(ULN-UIM)was built by considering the news geographical location,and the user interest model was built by combining ULN-UIM and UCN-UIM.The experimental results show that the performance of NCGP algorithm is better than that of collaborative filtering recommendation algorithm and UCN-UIM-based recommendation algorithm,and the accuracy of news event extraction algorithm can reach 93.6%.5.A news personalized recommendation prototype system was constructed.Firstly,the overall framework of the system was designed,including data layer,service layer and presentation layer.Then the four function modules of the system were designed.Finally,each function module was implemented.