Research On Key Technologies Of Information Source Localization And Hot Events Repository Of Online Social Networks

Hot social events often cause a lot of discussion in online social networks,which triggers wide spreading of relevant information in online social networks.The online social network plays an important role in the evolvement of the social events.The analysis of relevant information is essential for preventing the spreading of rumors and tracking events,which is based on the effective utilization of online social data.However,existing application systems cannot fulfill the requirements,and there are some key problems unsolved.The key problems in the process from raw online social data to the applications that can support decision making come from multiple aspects,and they affect each other.In order to solve the problems and support the analysis of relevant information of hot events in online social networks,the problems are studied in the framework of a hot event repository.Specifically,three problems are studied.Firstly,the lack of unified data model for heterogeneous online social data and corresponding data integration method hinders the effective utilization of online social data.Secondly,for source localization,which is vital in preventing the spreading of rumor,the lack of applicable source localization method obstructs the implementation of such application.Thirdly,since there are many applications and sets of data resource,it is hard to find the application and the data needed.In order to mitigate these challenges,in the framework of the hot event repository,the following research is performed.Firstly,aiming at the governance of the heterogeneous online social data,a unified data model for information diffusion analysis and source locating based on semantic web technologies and data modeling technologies is proposed.The representation and management of heterogeneous online social data is studied.This work is a good reference for applying algorithms and models to real online social data.In addition to the unified semantic model that describe the entities and relationships in the online social data,a schema mapping method based on instance-level similarity is proposed,which is designed for the schema mapping task when no property names can be used.Experiments show that for the property whose length of value varies more,the mapping is more accurate.Secondly,aiming at locating the source of asynchronous information diffusion process when the diffusion process cannot be fully observed,an estimator for source locating and a method to get the approximate diffusion time delay are proposed.Given a limited number of observable nodes,the source locating method is studied under different parameters of diffusion on different networks with different sampling strategies.The proposed method outweighs the-state-of-art method in precision and adoptability.Besides,the effect of parameters is studied.The proposed source locating estimator combined with the betweenness based sampling method is better than other combination.When the forward probability is smaller,the precision of source locating is lower.Experiments also show that the proposed method to approximate the diffusion time delay can improve the performance of the estimator.Thirdly,aiming at searching for data resources and analysis applications,the data resources and analysis applications are encapsulate as services,and a service discovery method based on semantic similarity is proposed.The method takes the input and output of requests and services as features for matching.The proposed service discovery method has three components:preprocessor,filter and matcher.The first two components reduce the range of searching.The relationship between concepts is taken into consideration in the design of the matcher.Experiments show that comparing to the method using a similar assumption and the proposed method has better precision and recall.Besides,the running time of proposed method is lesser in the same setup.Finally,the design of a prototype system is given.The feature comparison shows the advantages of this work.Demonstration and case study are provided.It provides some guidance for the design and implementation of online social network analysis application systems.