Kartik Mishra
Management
September 2024
Fake profiles have become the most significant problem due to rapid expansion of social media, threatening user privacy, security, and integrity online. This paper aims to improve the precision of false profile identification based on machine learning approaches used in identifying fraudulent accounts. Random Forest, XGBoost, and LSTM are the three machine learning models that are trained and evaluated. Specifically, 75 of the profiles in the MIB dataset are real, while the other 75 are false. This creates a balanced dataset of 150 profiles. The models are assessed using measures such as recall, accuracy, precision, F1-score, and ROC curve. The findings demonstrate that XGBoost performed better than the other models. Its accuracy was 98.7%, precision was 97.8%, recall was 99.0%, and the F1-score was 98.4%, indicating that XGBoost has the potential to detect fake profiles more effectively. It's, on the contrary, quite worse for LSTM, with an accuracy of only 89.3%, whereas the performance for Random Forest resulted at 93.3% accuracy. The confusion matrix and ROC curve analysis further enhanced XGBoost's outstanding performance through its lowest false positive rate and the greatest true positive rate. The study ends with showing how well the machine learning models, specifically XGBoost, detect fake social media profiles, with insights for enhancing social media security and fighting online fraud
121- 132
© Airo Journals 2021
