EXPLORING VARIOUS FEATURES FOR IDENTIFICATION OF FAKE PROFILES USING MACHINE LEARNING

Abstract

Instagram is gaining popularity among youth as it allows for self-expressiveness, promotion, and social connection. However, this popularity raises an issue of fake profile creation being used of disinformation, impersonation, and fraud, critical for user trust and the integrity of social networking platforms. In this research study, aim is to identify the fake Instagram profiles and classify them either fake or real. Focusing on extracting a comprehensive set of dataset features, such as user profile information, tagging and hashtags, content characteristics, interaction metrics, and content similarity measures, using the Fake Authentic Instagram User Dataset. Using the feature selection techniques, such as relief, correlation, gain ratio, and information gain, for ranking top features. For empirical analysis, a diverse range of conventional machine learning and deep learning techniques applied to evaluate their efficacy in classifying profiles. The algorithms that undergo the evaluation consist of shallow models, ensemble models deep learning models. The performance of each algorithm is estimated by standard performance evaluation measures of classification. The results show that Random Forest, an ensemble-learning model, outperforms other models with an accuracy of 95.10% in distinguishing between the fake and real Instagram users’ profile.