Novel Learning Approaches for Feature Representation and Classification Using Learning Approaches
T Dhakshnamoorthi, M. Giriraj, S. Gokul, K. Vanitha · 2024
While social networking sites such as Facebook, Twitter, and others offer many benefits, they also have many drawbacks. Cyberbullying is one problem these social networking sites have. Because it varies so much on the victim, the impacts of cyberbullying on their lives are incalculable. For victims, the message can be one of bullying, but for others, it might be expected. It is quite challenging to identify the bully material because of the ambiguity in cyberbullying texts. There have been reports of studies that solve this problem with text posts. On the other hand, less research has been done on text and image-based cyberbullying detection. Creating a model that will aid in preventing problems with image-based cyberbullying on social media platforms is the aim of this research. Initially, machine learning is employed for model building. Later, this study makes use of feature representation and classifier models. The experiment results using different hyperplane settings suggest that the model based on learning is the more suitable option for this particular assignment. In the optimal scenario, the suggested model's accuracy of 97.2% shows that the system can identify most posts involving cyberbullying. RMSProp approach appears to perform the best in terms of accuracy, sensitivity, specificity, and precision.