Prediction of Political Elections from Social Media Postings Mathematical Model of Hit Phenomena

Using the mathematical model for hit phenomena, we analyze the political elections in Japan. The mathematical model of hit phenomena is a theory of time-dependent humanhuman interaction based on the many-body theory in physics. In contrast to the simple measurement of the total number of posting to weblogs and twitter for each candidate of elections, we found that the direct communication strength and the indirect communication strength defined in the mathematical model for hit phenomena play significant role in the analysis of the daily number of posting to weblogs and twitter. The order of voting for each political elections in Japan can be reproduced exactly using our model for the Oaka mayor election in 2011 and the election of the Osaka area of the House of Councillors of Japan in 2010. Introduction Today, social media becomes very popular so that every social activities are affected at least partially from social media like weblogs, Twitter, Facebook and Google+. For economic activities, social media should be considered in marketing. For political activities, the effect of social media is also considered to be very strong not only the Arab Spring in 2010-2011 but many other countries in the world[1,2]. Especially for political elections, the effect of social media is very significant to predict the result of the elections before the pol. For the Japanese election we concern, there are some studies for the correlation between political election and social meda[3]. In this paper, we apply the mathematical model for hit phenomena to analyze the effect of social media to the political elections in Japan. Our approach can be used for the prediction of results of political elections. The mathematical model of hit phenomena we used here is presented by one of the author for the prediction of hit phenomena in entertainment businesses[4-9]. Hit phenomena in entertainments are very popular dynamical phenomena in real economics. For analysis of hit phenomena, we consider it as time-dependent non-equilibrium phenomena. In the mathematical model of hit phenomena, we calculate the purchase intention of each consumers as the result of the effect of advertisements, direct communication among consumers and indirect communication or rumor as an solution of differential equation for time. For the purpose of the present paper, we apply the similar equation for vote intention. In this study, we consider the hit phenomena of political votes with both experimental and theoretical ways. For experimental viewpoint, we observe daily weblog-posting data and twitter-posting data for several candidates in each Japanese political election using the Hottolink Co.Ltd, a social media monitoring platform vendor in Japan. We also obtain the daily exposure for each candidate on television using the database presented by M Data Co. Ltd. in Japan. For theoretical viewpoint, we present here a mathematical model for hit phenomena as microscopic voter model where we treat the TV exposure as an external force from the candidate side and the voter to voter interaction as many body interaction just like many-body theory in solid state physics. In the present model, equation of voter action is presented for each voter persons. Observation of election We observe three Japanese political elections; Osaka Mayor election in November 2011 and the election of Osaka and Tokyo prefectural constituencies of the House of Councillors of Japan in July 2010. The daily weblog posting data is monitored by using the social media monitoring platform "Kuchikomi Kakaricho" of Hottolink for these three elections. For the recent Osaka Mayor elections, the Twitter posting is also monitored. For the three elections, the exposure of election candidates on television is monitored by using the data set presented by M Data. The data of the weblog posting and the exposure data on television are collected about one month duration before the