Fraud News Detection for Online Social Networks

Abstract

Social media plays a vital role in all online aspects now, including personal communication, business and economics. It even affects political aspects seriously. A huge amount of available information, especially micro blogs is considered as a massive growth rate of human users, which is represented in the unprecedented diversity of its participants in terms of backgrounds, reasons and languages a revolution in its possibility of sharing public information, besides there is the way it makes its participants use their devices and perform their mission. 

Twitter, as a most famous used type of online social networking, contains huge data and news that throw the light on the content investigation in the tweets. This paper has discussed a proposed approach for determining the credibility of spread news on such social networks in two phases: The first phase is to detect the fake users enabling to ignore the news given by fake users. The second phase detects the credibility of the news content for the previously checked  

Account users by using the similarity measures and most popular machine learning algorithms such as (Support vector machine, Decision tree, Neural networks, Naive Bayes, Random forest) that enhance the credibility examining. The accuracy of the results of this phase is 99.8 %. In the second phase the news content credibility is detected by using the most popular similarity measures (Jacard, Cosine and Dice), which Jacard ended up with 95.4%percentage of accuracy.