Models Of The Storing Big Data In The Tasks Of Social Networks Monitoring

Nowadays,social networks are one of the main methods of communication and quick information exchange between people.Hereby social networks are understood as interactive multi-user web-sites aimed at building,reflecting and organizing interactions between users by uniting them on the relative characteristics or consolidating basing on common interests.Due to their structure,social networks contain a huge amount of dynamically changing data about users' interests,communication style,political trends etc.that influences the formation of the public opinion,business and promotion of goods and services.Thereby,they aggregate a huge amount of information exceeding millions of terabytes.Social networks monitoring is applied to use this information in various applied tasks.It allows to do behavioral modeling,to classify and define hidden unions and groups,to do clustering and filtering of certain users' characteristics in various researches.To solve such tasks there appears a need in processing big amount of information that can be done by applying Data Mining(DM)and Social Network Analysis(SNA)methods.The right choice of the technology for collecting,storing and processing big data should be made to implement these tasks.Moreover,a great attention is paid to the choice of the tools for storing big data since the efficient structure of data storage and extraction allows to optimize technologies of parallel data processing in the distributed environment.The research aims at solving the problem of choosing the complex of hardware,software and efficient algorithms of processing big data in near real-time conditions in the tasks of social networks monitoring by creating a universal model of storing big data.The author analyzes and synthesizes the meta-model of storing big data that includes mathematical models(to solve direct and inverse problems of quantitative evaluation of big data characteristics)and data storage models(to save different categories of social networks data at the computing resource).The conducted experiments demonstrate that the suggested model allows to reduce data receiving time and make it near real time by loading data,that is actual at the moment of analysis,for processing into the computing resource.The theoretical and practical results received in the Master's theses can be used by big data analysts while developing software for social network analysis.