Identifying fraud content within social-media using naive bayes algorithm compared over XGboost algorithm with improved accuracy

• Main Authors: Palaparti, Ikya, Amanullah, M.
• Format: Conference Proceeding
• Language: English
• Subjects: Accuracy; Algorithms; Datasets; Digital media; Fraud; Machine learning; Motion perception; Social networks; Supervised learning
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0232784

Improving the accuracy of social media scam content identification is our primary motivation for doing this study. We launched this programme to help detect false material published on social media, which is becoming more important as the volume of fake news continues to rise. By implementing a number of interrelated safeguards, fraud detection aims to stop the fraudulent movement of money and other assets. Research Methods and Equipment: Using unique naive bayes and XG boost with variable training and testing splits, we are able to predict and identify social media fraud content. A whopping 80% is the gpower. With α=0.05 and power=0.80, the Gpower test produces a result of around 85 percent. Using the classification schemes described here, it should be easy to spot publications that aren’t based on this principle. This approach uses a new naive bayes algorithm to categorise the dataset. At its core, this initiative is concerned with the political online source dataset. Messages are categorised as either trustworthy or fraudulent in this new benchmark dataset for spam identification. We have already looked at the "Liar" dataset. The confusion matrix displays the outcomes of the dataset analysis performed using the five approaches, as shown by a 2-tailed significance value of p=