부산대학교 도서관

Social media networks are now essential and play a significant role in society. According to data, the number of active users on various social media platforms including Facebook, WhatsApp, Instagram, and many more is growing rapidly. As a result, there is an increase in risky actions, making the area more unsafe. Personal information security is now seriously threatened. The search for anomalous users is a field that is constantly being researched, but because of the threat that it poses, it is also a field that will never end and will face numerous obstacles, including accuracy. Different Machine Learning and Deep Learning models have been proposed and created by numerous researchers. But, many of these models have scope for improvements, in terms of accuracy and reducing false positives, reducing false negatives. To achieve these enhancements, we have compared different models and using our hybrid model, with attempts for increasing accuracy. In this research we have implement an accuracy-based model named GridBoost which uses hyperparameter parameter tuning fusion with XGBoost. We used a variety of popular classifier models, including Linear Regression (LR), Naive Bayes (NB), KNN (K-Nearest Neighbor), Support Vector Machine (SVM), and GridBoost, which were developed for anomaly identification using four different standard datasets. The performance study shows increased accuracy with our proposed hybrid technique up to 98% when compared to other assessment metrics like precision, recall, and F1-score. [ABSTRACT FROM AUTHOR]