EMOTION-DRIVEN CYBERBULLYING DETECTION WITH LSTM–GRU ENSEMBLE AND SECURE FLASK-BASED DEPLOYMENT

Renuka Chowdary Bheemana; SAGGURTHI BINDU SARVANI; PALADUGU MAHITHA; MUSUNURU SYAM PRASAD

doi:10.62643/

Authors

Renuka Chowdary Bheemana Author
SAGGURTHI BINDU SARVANI Author
PALADUGU MAHITHA Author
MUSUNURU SYAM PRASAD Author

DOI:

https://doi.org/10.62643/

Keywords:

Cyberbullying Detection, Emotion Analysis, Sentiment Features, LSTM, GRU, Ensemble Deep Learning, BERT, Toxic Dataset, Twitter Dataset, Sequential Modeling, Flask Front-End, User Authentication, Real-Time Prediction, Natural Language Processing (NLP).

Abstract

Cyberbullying detection has evolved with the integration of contextual, emotional, and sentiment-based features; however, further improvements are required for real-time and highly accurate predictions. In this extension, we enhance the existing cyberbullying detection framework by incorporating advanced ensemble deep learning models such as LSTM and LSTM+GRU, which efficiently capture sequential patterns in user-generated text. These models significantly improve classification performance and achieve up to 99% accuracy, demonstrating their effectiveness in detecting complex cyberbullying behaviors. To support practical usability, a Flaskbased front-end system with secure authentication is developed, enabling users to interact with the detection system in a simple and secure manner. The extended system offers improved accuracy, stronger robustness, and enhanced accessibility, making it a highly suitable solution for real-world cyberbullying monitoring and intervention applications.