Identifying misinformation on Twitter with a support vector machine

摘要

There is a large amount of information from disparate sources around the world. Due to the recent growth of online social media and its impact on society, identifying misinformation is an important activity. Twitter is one of the most popular applications that can deliver engaging data in a timely manner. Developing techniques that can detect misinformation from Twitter has become a challenging yet necessary task. This article proposes a machine learning method that can identify misinformation from Twitter data. The experiment was carried out with three widely used machine learning methods, na?ve Bayes, a neural network and a support vector machine, using Twitter data collected from October to November 2017 in Thailand. The results show that all three methods can detect misinformation accurately. The accuracy of the na?ve Bayes method was 95.55%, that of the neural network was 97.09%, and that of the support vector machine 98.15%. Furthermore, we analyzed the misinformation and noted some of its characteristics.