ตัวแบบเชิงลึกเพื่อตรวจจับข่าวปลอมภาษาไทยบนสื่อสังคมออนไลน์

Abstract

Nowadays, fake news on social media has caused many problems because they spread easier and faster than the real ones, while fake news detection or examination consumes high resources (human power, time, etc.). Thus, there is a need for an automatic method to examine or verify, so this research aims to find significant features of fake Thai news and an appropriate machine learning model between Decision tree, Support Vector Machine and Neural Network model to examine the fake Thai news on Twitter. The evaluation results show that the significant features of fake Thai news are the amount of follower, the sentiment score of news content, the length of content’s character, the amount of retweet, the ratio of friend and follower, the amount of news favorited, the amount of post since signing up. The machine learning model that suits to examine the fake Thai news is a Neural Network model which performs 97 percent of accuracy.