Identification of Spambots and Fake Followers on Social Network via Interpretable AI-Based Machine Learning

The identification of spambots and fake followers on social media platforms is an essential task to ensure the authenticity of user interactions and maintain the integrity of online environments. Social networks are increasingly affected by the presence of fake accounts that inflate user engagement metrics, spread misinformation, and disrupt meaningful conversations. To address this issue, this project proposes a solution that employs machine learning algorithms combined with interpretable AI techniques to identify and classify spambots and fake followers. The dataset used for training the models contains various features such as user interactions, post content, engagement metrics, and other metadata. The project utilizes K-Nearest Neighbors (KNN), BaggingClassifier, StackingClassifier, and CatBoostClassifier for the task of classification. The interpretability of the models is enhanced using Partial Dependence Plots (PDP), which provide valuable insights into how different features influence the prediction outcomes. Model evaluation is conducted using metrics like accuracy, precision, recall, and F1-score, ensuring that the models perform optimally in distinguishing fake accounts from legitimate users. This project aims to develop an automated solution that aids in the identification of fake followers and spambots, offering insights that enhance platform credibility and user trust. The results of this work can contribute to the development of more transparent and efficient methods for detecting fake activities across various platforms.

Keywords: Spambots, Fake Followers, Machine Learning, Classification, Partial Dependence Plots, CatBoost, BaggingClassifier, K-Nearest Neighbors, StackingClassifier, Web Interface.