A comparative analysis of machine learning algorithms for hate speech detection in social media

Abstract

A detecting and mitigating hate speech in social media, particularly on platforms like Twitter, is a crucial task with significant societal impact. This research study presents a comprehensive comparative analysis of machine learning algorithms for hate speech detection, with the primary goal of identifying an optimal algorithmic combination that is simple, easy to implement, efficient, and yields high detection performance. Through meticulous pre-processing and rigorous evaluation, the study explores various algorithms to determine their suitability for hate speech detection. The focus is finding a combination that balances simplicity, ease of implementation, computational efficiency, and strong performance metrics. The findings reveal that the combination of naïve Bayes and decision tree algorithms achieves a high accuracy of 0.887 and an F1-score of 0.885, demonstrating its effectiveness in hate speech detection. This research contributes to identifying a reliable algorithmic combination that meets the criteria of simplicity, ease of implementation, quick processing, and strong performance, providing valuable guidance for researchers and practitioners in hate speech detection in social media. By elucidating the strengths and limitations of various algorithmic combinations, this research enhances the understanding of hate speech detection. It paves the way for developing robust solutions, creating a safer, more inclusive digital environment.