Improvement performance by using Machine learning algorithms for fake news detection

Abstract

The prevalence of internet use and the volume of actual-time data created and shared on social media sites and applications have raised the risk of spreading harmful or misunderstanding content, engaging in unlawful activity, abusing others, and disseminating false information. As of today, some studies have been done on fake news recognition in the Kurdish language. For extremely resourced languages like Arabic, English, and other international languages, false news detection is a well-researched research subject. Less resourced languages, however, stay out of attention because there is no labeled fake corpus, no fact-checking website, and no access to NPL tools. This paper illustrates the process of identifying fake news, using two components of the dataset for fake news and actual news. Several classifiers were then applied to the quantity after using identifiers as a highlight of selection. Results of the proposed study demonstrated that Passive-Aggressive Classifier (PAC) outperformed the other classifiers on both datasets the dataset with an accuracy score of 93.0 percent and other classifiers were less in some percentage that show high accuracy as well since it is 90 percent.