TWITTER SENTIMENTAL AND EMOTIONAL VARIATIONS OF PUBLIC REACTIONS ABOUT THE PUBLIC PERSONALITIES IN SOCIAL MEDIA USING DEEP LEARNING

Abstract

It is difficult to predict the personality of people with high accuracy but it is possible nowadays by using the information which are shared by them in the social media. In order to find the personality with high accuracy, we examine Twitter account of six categories of the Public. Based on position in the society, we categorized the people into six major categories such as Political Personalities, Sports Stars, Business Bodies, Hollywood Figures, General Public and Indian Geneticist. Their twitter profile data are collected, stored (using MySQL), Pre-processed and Analysed analysis is done in accordance with the features of profile images such as colour, image composition, their Post and their online engagement. For explicable, our analysis targeted aesthetic and physiognomy to explore personality characteristics. Based on the profile picture chosen our results predict the behaviour of the twitter users. For example, studious and agreeable users having positive emotions while others use aesthetic photos. After analysing these data, the accuracy rate was predicted by using various algorithms such as Support Vector Machines (SVM), Naïve Bayes and Maximum Entropy etc. These algorithms gave the accuracy rate of about 60-70%. For accuracy prediction we used D-CNN technique, which achieved 98.457% of accuracy measurement and also concluded that personality prediction is difficult for robust accuracy.