Predicting the Outcomes of Important Events based on Social Media and Social Network Analysis

Twitter is a famous social network website that lets users post their opinions about current affairs, share their social events, and interact with others. It has now become one of the largest sources of news, with over 200 million active users monthly. It is possible to predict the outcomes of events based on social networks using machine learning and big data analytics. Massive data available from social networks can be utilized to improve prediction efficacy and accuracy. It is a challenging problem to achieve high accuracy in predicting the outcomes of political events using Twitter data. The focus of this thesis is to investigate novel approaches to predicting the outcomes of political events from social media and social networks. The first proposed method is to predict election results based on Twitter data analysis. The method extracts and analyses sentimental information from microblogs to predict the popularity of candidates. Experimental results have shown its advantages over the existing method for predicting outcomes of politic events. The second proposed method is to predict election results based on Twitter data analysis that analyses sentimental information using term weighting and selection to predict the popularity of candidates. Scaling factors are used for different types of terms, which help to select informative terms more effectively and achieve better prediction results than the previous method. The third method proposed in this thesis represents the social network by using network connectivity constructed based on retweet data and social media contents as well, leading to a new approach to predicting the outcome of political events. Two approaches, whole-network and sub-network, have been developed and compared. Experimental results show that the sub-network approach, which constructs sub-networks based on different topics, outperformed the whole-network approach.