Improving Fake Profile Detection: A Hybrid Machine Learning Approach with Negative and Clonal Selection

Abstract

The spread of fake profiles on social networking websites is a growing problem, posing significant challenges to user safety and data integrity. This research addresses the critical need for efficient detection mechanisms by proposing a hybrid machine learning model that combines negative selection and clonal selection algorithms. The study employs three different datasets, including two from Instagram and Twitter, following a detailed methodology for data preprocessing, feature extraction, and model implementation. Negative selection is applied to filter out irrelevant samples, while clonal selection enhances the model by optimizing solution discovery. The results show that the hybrid approach significantly improves detection accuracy, with a Random Forest model achieving an impressive 99% accuracy. Precision, recall, and F-measure tests further demonstrate the superiority of this new method over traditional techniques. The outcome of the research determines the efficacy of using negative and clonal selection algorithms together to detect fraudulent profiles more efficiently and maintain digital integrity. The research forms the foundation for additional research in this field.