Sentiment Analysis of the Conflict Between Meitei and Kuki Communities of Manipur, India Using YouTube Comments: A Comparative Study of Random Forest, Naïve Bayes, SVM, and KNN

Purpose: The primary goal of the study is to use YouTube as a data source to examine public opinion about the ethnic conflict or tension between the Meitei and Kuki communities of Manipur, India. It investigates how digital platforms reflect emotional engagement with regional conflict and evaluates machine learning models for sentiment classification in this context. Methodology: A dataset of 551 YouTube videos and 124,971 associated comments were collected through Webometric Analyst and preprocessed using natural language processing techniques, including TF-IDF vectorization. Four models for machine learning to classify comments into positive, neutral, or negative attitudes, Support Vector Machines (SVM), Random Forest, Naïve Bayes, and K-Nearest Neighbors (KNN) were trained and evaluated. The model's performance was evaluated using its F1-score, recall, accuracy, and precision. Findings: SVM achieved the highest accuracy (93.42%), followed closely by Random Forest (91.59%). Many comments were neutral, while positive sentiments marginally exceeded negative ones. The sentiment distribution suggests a complex emotional response, including empathy and calls for peace. The top 10 most-viewed videos revealed various content types, from news and interviews to political commentary and vlogs, indicating diverse user engagement with the issue. Originality: This study presents one of the first applications of machine learning-based sentiment analysis to YouTube content on the Manipur conflict. It contributes original insights to computational social science and digital media research by demonstrating how user-generated video content can assess public sentiment in real time. The findings offer methodological and substantive value, particularly in understanding how digital discourse shapes perceptions of ethnic conflict in India.