Research Paper | Computer Science | Indonesia | Volume 9 Issue 5, May 2020
Incremental Machine Learning to Detect Fake New Using Support Vector Machine
Edwin F Wicaksono, Ruddy J Suhatril, Matrissya Hermita
As one of the biggest online market worldwide, 97.4 % of internet users in Indonesia are active on social media . Facebook, one of the most common social media in Indonesia, provides a platform to spread information throughout its user. However, fact-based information is not the only one circulate on the Internet. The number of fake news shared throughout the Internet, especially social media, is concerning. Investigating fake news requires considerably longer time in collecting the data to compare. In addition, humans naturally are not very good at differentiating between real and fake news . It makes machine learning becomes advantageous in dealing to this problem. However, the rapid changes of news throughout time requires machine learning to be able to train its model dynamically. Incremental machine learning is proposed to solve this problem. As much as 6757 labeled data containing both fake and factual news provided by George McIntire, Politifact, and Buzzfeed are set to be primary data in this study. In addition, over 30.000 crawled news from various reliable sources are prepared to observe the most efficient data ratio to train the model. Based on model-selection approach, Support Vector Machine outperformed the other models with the initial accuracy of 0.889. Along with feature extraction, parameter tuning, and feature selection, performance of the incremental machine learning can reach over 96 % accuracy.
Keywords: Machine Learning, Fake News, Support Vector Machine
Edition: Volume 9 Issue 5, May 2020
Pages: 666 - 672
How to Cite this Article?
Edwin F Wicaksono, Ruddy J Suhatril, Matrissya Hermita, "Incremental Machine Learning to Detect Fake New Using Support Vector Machine", International Journal of Science and Research (IJSR), https://www.ijsr.net/search_index_results_paperid.php?id=SR20510130034, Volume 9 Issue 5, May 2020, 666 - 672