Research On Key Techniques Of Information Precise Service In Military Information Utility Process

With the constant improvement of new military reform, the syncretism trend of computer network, communication network and command and control network has been very clear for everyone. Besides, military information in these networks has presented the many new characteristics like very large scale, real time and illusive. Thus, how to find the useful information in very large scale data has been the key problem in military information utility nowadays. In order to distill and dig the useful information and upgrade the capability of information utility, military information precise service has been considered as the key for solving the problem of military information utility.This dissertation takes military information utility as research background, takes information precise service key techniques as research content, puts emphasis on the challenges of military information utility, and proposes the addressed methods which can be summarized as follows:(a) military information character capturing method;(b) military information recommendation method;(c) users’ relationship analysis and prediction method;(d) military information utility mode. In the above research contents, this dissertation mainly focuses on solving some hot and hard problems of information utility which both industry and academy focus on, such as “cold start” problem in information recommendation, data redundancy detection, users’ social relationship analysis and process reservation problem, and gives the essential technical support for military information precise service.(1) The high dimensions of military information will greatly increase the complexity of information utility and impact the performance of military information precise service. Aiming at the characteristics of high dimensions and high complexity for military information, this dissertation proposes the methods of military information character capturing. This dissertation definitude the concept and connotation of military information character capturing and systemically introduce the process of military information character capturing. This dissertation proposes the concrete approach based on text analysis, image analysis, video analysis and audio analysis methods. Besides, this dissertation also detailed analyzes the military task species and then proposes the task character capturing methods. At the meantime, this dissertation proposes user character capturing approach based on the analysis of users’ background information and visit information. Finally, this dissertation proposes the similarity matching algorithm between information character and users’ demand, based on which we get the results that proper information can be recommended to the proper users.(2) With the fast development of network technology and information science, the very large scale information utility has been considered as one of the most critical problems faced by users. This dissertation proposes the methods of military information recommendation from two aspects to solve the problems of “cold start”, data redundancy detection and dynamic military information and users’ preference:(a) information recommendation based on the hybrid recommendation algorithm and hidden markov model;(b) information recommendation based on time dynamics modeling. When studying the content of “information recommendation based on the hybrid recommendation algorithm and hidden markov model”, this dissertation detailed analyzes the traditional content based recommendation and collaborative filtering method, and proposes the improved hybrid method which both utilizes the advantages of user-based collaborative filtering and item-based collaborative filtering methods. It fulfills the requirements of the scalability and automatism of information recommendation. This dissertation also propose a recommendation method for new user based on the popularity and entropy techniques, using which we capture the preference of new user and solve the “cold start” problem better than the traditional methods. Besides, this dissertation proposes a data redundancy detection method based on the hidden markov model analysis. It synthetically analyzes the trend of information change and finds the way to identify the redundant data, and then we solve the problem of redundancy detection at a certain extent. When studying the content of “information recommendation based on time dynamics modeling”, this dissertation proposes a method to find latent friends of users based on the interest-similar cluster identification, using the information of users’ groups, group scale, common group numbers and common tags. This dissertation also proposes a time dynamic model of user behavior and builds the relationship among interest-similar cluster, user and information. Based on the relationship built above, we recommend the proper information to users using the unbiased random walk algorithm and get the good results that people can find the favorite information she does not know before, the experiments based on the real datasets also prove that our approach can get better recommendation results than the traditional methods.(3) In the Web 2.0 era, various social Medias which have provided much help in users’ communication and information utility have been considered as a very important factor in military information precise service. In order to make the best of the users’ relationship in information utility, this dissertation proposes a method to predict and analyze the relationship of users. This dissertation first proposes a node importance evaluation method based on node centrality evaluation, Pagerank and HITS algorithms to predict the importance of users. In the condition that almost every social network only shows the neighborhood list out to the users, this dissertation proposes a social relationship identification method only based on topological information analysis. The experiments show that it gets a very good performance in identifying the relationship existence. As there exist several of relationship types between users such as friend and enemy, a relation type identification approach is studied. It improves the relationship prediction performance and identifies the key characters in relationship prediction process.(4) With an increasing use of services in information utility era, ensuring optimal use of resources to satisfy user’s preferences becomes crucial. This dissertation proposes an effective military information utility mode based on process reservation. It first builds the architecture of process reservation, and then does the four major jobs: service differentiation, service reservation, process reservation and Qo S control. It considers the application process as a whole instead of considering single service reservation success as the aim, so it increases the success rate in the process level. It is implemented from two aspects: abstract process reservation and concrete process reservation. Besides, this dissertation proposes a reservation method comprehensively using the relationship of services for better reservation and a policy-based reservation method for flexible reservation. The comprehensive experimental analysis shows that our approach achieves better results both in users’ flexibility and resource utilization.