Sentiment Analysis of Telegram App Reviews Using Text Classification Models

This study explores sentiment analysis of Telegram app reviews using four machine learning-models: Naive Bayes, Random Forest, Logistic-Regression and Neural Networks. A dataset of 76,500 reviews was analyzed, and the models were evaluated based on precision, recall, F1-score, and accuracy across three sentiment categories: Negative, Neutral, and Positive. Among the models, random -forest achieved the highest accuracy of 75% but struggled with Neutral sentiment classification, a challenge shared by the other models. Logistic Regression and Naive Bayes both delivered 74% accuracy, with slightly better precision and recall for Negative and Positive sentiments. Neural Networks achieved 72% accuracy but faced difficulties in identifying Neutral sentiments. The use of “Natural Language Processing” (NLP) techniques, such as text preprocessing and feature extraction, proved crucial for improving performance.