Comparative Performance Evaluation of a Proposed AI-Based Text Mining Framework for Human Sentiment Analysis Using Reliability and Accuracy Metrics

With the proliferation of social networks, a tremendous amount of content has been created by users, making sentiment analysis one of the most crucial tools for understanding user opinions and emotions. But this is a critical challenge in getting reliable sentiment from noisy and unstructured social media data. The current research builds a comprehensive analysis framework of human sentiment analysis machine learning models based on standard reliability and accuracy measures. Logistic Regression, Naïve Bayes, Decision Tree, Random Forest and Support Vector Machine are compared within the same frame work under the same experimental conditions. The accuracy, precision, recall, F1-score, Cohen's Kappa, Matthews Correlation Coefficient and error rate are used to measure the performance. The experimental results indicate that the model with Random Forest approach has the highest classification correctness (34.01%) and reliability score (0.3275), indicating that the model has better prediction performance and stability in the classification of social media sentiment