Methods of Exposure Informatively Psychological Influences in Social Networks

Abstract

Purpose of the article is creation of the system of analysis of data of method of exposure of trolingu by determination of the key of text kontentu of social networks; receipts of indexes, which characterize the presence of signs of trolingu in text; calculation for these indexes of informative entropy of text kontentu and comparison of it numerical value with possible maximum. Research methods are analysis of accordance of the analytical systems the requirements pulled out to them and exposure of advantages of neyromerezhevogo approach by comparison to traditional mathematical methods: mathematical statistics, cluster, regressive, factor analysis. As an intellectual system of data analysis, that most satisfies modern requirements to the analytical systems, in this work it is suggested to choose the class of the systems with the use of technology of neyromerezhevogo analysis. The novelty of the conducted research is a construction of artificial neyromerezhi, which owns the followings advantages: high-efficiency paralel’no-poslidovna treatment of information, maximal potential parallelism and most effective use of any parallel calculable architecture, and it is in comparing to other calculable technologies. Conclusion. Mass parallelism of neyroobchislen’, necessary for effective treatment of appearances, it is provided locality of treatment of information in neyromerezhakh. Every neuron reacts only on local information which acts to it presently from pov’’yazanikh with it the same neurons, without an appeal to the general plan of calculations. Thus, neyromerezhevi algorithms local, and neurons are able to function parallell.

Keywords

• нейрон
• нейронна мережа
• аналіз даних
• глибоке навчання
• тролінг
• тональність текстового контенту