An optimized Arabic cyberbullying detection approach based on genetic algorithms.

Area of Science:

• Natural Language Processing
• Computational Linguistics
• Social Computing

Background:

• Cyberbullying poses a significant threat on digital platforms, causing psychological harm.
• Detecting cyberbullying in Arabic is challenging due to dialectal variations, informal language, and contextual nuances.
• Conventional tools struggle with the complexities of Arabic cyberbullying identification.

Purpose of the Study:

• To enhance Arabic cyberbullying detection mechanisms.
• To introduce a feature-selection strategy using a Genetic Algorithm (GA).
• To improve the accuracy and efficiency of identifying harmful online content in Arabic.

Main Methods:

• Utilized a Genetic Algorithm (GA) for feature selection on a corpus of 46,000 Arabic Instagram comments.
• Applied GA to reduce feature space, preserving semantic structures and removing noise.
• Evaluated four classifiers with GA-driven feature selection.

Main Results:

• The GA feature selector reduced the feature space by approximately 50%.
• GA-driven selection improved F1-scores by 3.45-14.96% across classifiers.
• Classification time was reduced by a factor of 2.32-12, indicating significant efficiency gains.

Conclusions:

• Genetic-feature optimization enhances precision and significantly improves runtime for cyberbullying detection models.
• The proposed method enables scalable and context-sensitive detection of harmful language in Arabic.
• This approach is beneficial for morphologically rich languages requiring nuanced content analysis.