Analysis of Thai Fake News Using Naïve Bayes Models

Detecting fake news in an early stage, even though it is challenging, is thus adventurous to protect harms to people. In this paper, we present a framework for revealing the evidences of fake news in Thai news titles using the Naïve Bayes model. The framework enables us to discover footprints for fact news and fake news through the four back-to-back steps: data acquisition, data pre-processing, data exploration, and data modeling. We also intensely examine the Naïve Bayes model discrimination of fact and fake news when employing different text normalization methods. The experiments show that the Naïve Bayes model can achieve the accuracy performance up to 89%. Moreover, we provide the extensive discussions about our data exploration. We also discuss our application of posterior probabilities to reveal evidences of fake news.