Detecting Deceptive Identities: A Machine Learning Approach to Unveiling Fake Profiles on Social Media

In the digital age, online security and authenticity face serious issues due to the widespread use of fraudulent personas on social media sites. This research article outlines a comprehensive methodology for detecting fake profiles, starting with the gathering of user profile data from diverse social media sites. Pertinent characteristics such as completeness markers, posting frequency, network diversity, and language patterns are extracted from profiles to distinguish between real and fraudulent accounts. This work involves assessing various classifiers suitable for fake profile classification task, including Decision Trees (DT), Random Forest (RF), Logistic Regression (LR), Support Vector Machines (SVM), Adaptive Boosting (AdaBoost), Gradient Boosting (GB) and Voting Classifier (VC). The effectiveness of Principal Component Analysis (PCA) in improving model performance and reducing dimensionality is also investigated. Each classifier is trained on the Kaggle dataset after thorough evaluation, exploring both the raw features and those reduced by PCA. The Random Forest with PCA has the highest prediction rate in detecting fake profiles at 92.5%. The article concludes with insights into classifier performance and suggestions for further research to bolster fake profile detection accuracy and online community security.