<![CDATA[Cyberbullying is becoming an increasingly troubling issue in today's digital age, with serious impacts on the well-being of individuals and society as a whole. With the number of social media users continuously rising, there is an urgent need to develop effective solutions for detecting cyberbullying. This urgency negatively affects the well-being of individuals, especially children and adolescents. The Big Data era also brings many new challenges, including the ability of organizations to manage, process, and extract value from available data to generate useful information. The aim of this research is to develop Cybersentinel, a cyberbullying detection application that combines Machine Learning and VADER Lexicon approaches to improve classification accuracy. It involves comparing several Machine Learning algorithms optimized using the GridSearchCV technique to find the best combination of parameters. The dataset used consists of social media comments labeled as bullying and non-bullying. The successfully developed model uses the Support Vector Machnine algorithm, achieving a best accuracy of 98.83%. The system is developed using Python with the Streamlit framework. This application development follows the Design Science Research (DSR) approach, which integrates principles, practices, and procedures to facilitate problem-solving and support the design and creation of applications. Testing is conducted using blackbox testing. The results show that parameter optimization using GridSearchCV can significantly enhance model performance, and applying the DSR method allows for the development of Cybersentinel tailored to specific needs. Thus, Cybersentinel provides an effective solution for detecting cyberbullying and contributes to improving the safety of social media users.]]>
