The Internet’s global reach has revolutionized communication but has also given rise to online trolls, who employ disruptive tactics, harassment, and misinformation to disrupt online discourse. Detecting and mitigating troll behavior is a critical challenge for maintaining the quality and safety of online communities. The rapid proliferation of online communities and social media platforms has made the issue of troll behavior a pervasive and challenging problem. This paper focuses on different machine learning techniques to detect troll data in online social media. Hence it is crucial for maintaining the integrity and safety of online spaces. The paper discusses various models and analyze textual data, that can be obtained from various internet sources, including social media, review sections etc. Here we have applied different machine learning models along with different techniques then we compare the results and got to know the benefits of balancing the dataset, vectorizing etc.