Comparison of a Combined Model (K-Nearest Neighbor Algorithm and Support Vector Machine Algorithm), K-Nearest Neighbor Algorithm, and Support Vector Machine Algorithm to Detect Hate Speech on Social Media

Abstract: In this study we delve into the use of a combined machine learning model to counter the growing problems of hate speech on social media. Hate speech can significantly impact individuals, communities, and society. To counteract this, we have proposed a combined K-Nearest Neighbor (KNN) algorithm with a Support Vector Machine (SVM) algorithm called the KNN-SVM. We used a publicly available dataset from Kaggle [18]. The dataset contains labels such as hate speech, offensive language, and normal text. Our aim is to improve the classification metrics for accuracy, precision, recall, and F1-score on detecting hate speech. Then we compared our model’s performance to the individual KNN and SVM model. While all models faced significant challenges due to the imbalanced data with fewer hate speech examples, KNN-SVM achieved a promising result against the other models. It demonstrated the highest recall and F1-score for hate speech detection, indicating its effectiveness in identifying these crucial instances. However, limitations exist. The model’s overall accuracy was lower than the individual SVM model. Further research is needed to enhance the model’s effectiveness through data refinement techniques like balanced labeling and exploring advanced algorithms like deep learning.