Improving fake instagram profiles detection by features reduction

Fake profiles represent a growing online threat and are linked to a range of harmful activities, including fraud and misinformation. Machine learning algorithms are used for fake profile detections based on the dataset features clustering. This paper aims to improve the performance of the classification of fake Instagram profiles based on reducing features using the Expectation Maximization (EM) algorithm. The IJECE dataset has 18 features that will be reduced to 12 using the EM algorithm. The study evaluates the dataset via multiple classification methods. The Decision Tree (DT) performs well after applying the features reduction as the processing time reduces by 64% and the accuracy rarely changes.