心得分享

09/01/2019 專題報告
名稱	專題報告
描述	專題報告
日期	09/01/2019
課程名稱	專題實作
指導教師	陳惇凱教授

The prevalence of the Internet has amplified the propagation of fake news with people freely sharing information online paying no heed to consequences. Many were brought to awareness about fake news during the period of the 2016 US presidential election. Masses of fake news and clickbait about the two candidates were flying around at the time. Not only an epidemic in America, it is everywhere the Internet can reach. Just this year in India, a viral video of a child being kidnapped, later proven fabricated, was widespread throughout WhatsApp instigating attacks on strangers, eventually leading to homicide. Incidents such as these caused us to wonder whether we can do within our field of profession to help alleviate this issue. Therefore, the idea of building a fake news detection program using machine learning came to be.

In this project, our contributions are two folds. First, we did some researches and literature reviews into other people’s works. Second, we built our machine learning program to investigate whether machine learning could detect fake news. Building around the arch question of our project: “whether fake news can be tackled using machine learning”, we used the famous “bag-of-word” approach to extract features and then tested them on multiple classic text classification algorithms and a simple neural network model. We then selected the classifier that outperformed other classifiers, with an accuracy rate as high as 91.52%, lay out several hypotheses, and made a comparison of our experimental results with other people’s works. However, even though our classifier performed well with the assigned dataset, the classification capability dropped in performance when being tested with real-world data. Although that being the case, we still think that with a more sophisticated approach machine learning can be a more promising tool in helping humans to detect fake news.

更新日期：2019/1/9 下午 11:01:34