An Effective Machine Learning Model for Scam Profile Identification on Instagram using Boosting Classifiers and Neural Networks

Scammers are now common on social media sites like Instagram, which detect unsuspecting information about people. While AI has been found to contribute to the spread of misinformation, it can also help solve this problem through machine learning (ML). In specifically, this research examines Instagram's use of ML algorithms for the purpose of detecting fake accounts. The assessment uses the Fake/Authentic User Instagram dataset with 65,326 samples organized through 18 features to classify users as authentic or fake by employing advanced ML techniques. A complete preprocessing method was applied, including null data management, feature normalization, dataset balancing, label transformation, and Chi-squared testing for feature selection. Among the developed models, LightGBM achieved the highest performance, with an accuracy(acc) of 98.42 %, precision (prec) of 97.80 %, recall(rec) of 98.86 %, and an F1 score(f-measure) of 98.33 %. A comparison with baseline models like RF and XGBoost revealed that the suggested model performed better. In addition to providing insightful information for creating automated techniques to identify and stop online fraud, this work advances expanding subject of social media forensics.