Abstract

In the contemporary landscape of pervasive social media usage, the identification of fake profiles stands as a crucialelement in preserving online security. This research delves into the realm of machine learning algorithms, focusing on thecomparative analysis of LightGBM and SVM for the task of detecting fake accounts on social media platforms.Our study harnesses the power of these two algorithms, both known for their robust capabilities, and evaluates their performanceagainst each other. The research utilizes a diverse set of features derived from user profiles, posting behavior, and linguistic patterns,with Natural Language Processing (NLP) techniques applied to extract nuanced insights from textual content.The outcomes of our investigation reveal that LightGBM and SVM exhibit superior accuracy in the identification of fake profiles.Further exploration involves fine-tuning hyperparameters, optimizing feature engineering, and experimenting with ensemblemethods to enhance the overall efficacy of the models.As the social media landscape continually evolves, our findings contribute valuable insights into the strengths of Light GBM andSVM in addressing the dynamic challenges of fake profile detection. This research not only provides a foundation for practitionersto make informed decisionsin implementing detection systems but also offers a pathway forfuture advancements in refining machinelearning models to combat emerging threats in the realm of online social interactions.