| Autor: |
Ammar Almomani, Khalid Nahar, Mohammad Alauthman, Mohammed Azmi Al-Betar, Qussai Yaseen, Brij B. Gupta |
| Jazyk: |
angličtina |
| Rok vydání: |
2024 |
| Předmět: |
|
| Zdroj: |
International Journal of Cognitive Computing in Engineering, Vol 5, Iss , Pp 14-26 (2024) |
| Druh dokumentu: |
article |
| ISSN: |
2666-3074 |
| DOI: |
10.1016/j.ijcce.2023.11.002 |
| Popis: |
Cyberbullying detection on social media platforms is increasingly important, necessitating robust computational methods. Current approaches, while promising, have not fully leveraged the combined strengths of deep learning and traditional machine learning for enhanced performance. Moreover, online content complexity requires models that can capture nuanced contexts beyond text, which many current methods lack. This research proposes a novel hybrid approach using deep learning models as feature extractors and machine learning classifiers to improve cyberbullying detection. Extracting features using pre-trained deep learning models like InceptionV3, ResNet50, and VGG16, then feeding them into classifiers like Logistic Regression and Support Vector Machines, enhances understanding of the complex contexts where cyberbullying occurs. Experiments on an image dataset showed that combining deep learning and machine learning achieved higher accuracy than using either approach alone. This novel framework bridges the gap in existing literature and contributes to broader efforts to combat cyberbullying through more nuanced, context-aware detection methods. The hybrid technique demonstrates the potential of blending deep learning's representation learning strengths with machine learning's sample efficiency and interpretability. |
| Databáze: |
Directory of Open Access Journals |
| Externí odkaz: |
|
