Evaluation of the Effectiveness of SMOTE and Random Under Sampling in Emotion Classification of Tweets

This study evaluates the effectiveness of two sampling techniques, SMOTE (Synthetic Minority Over-sampling Technique) and Random Under Sampling (RUS), in improving the performance of several classification models, namely Maximum Entropy, SVM, Random Forest, Neural Network, and Naive Bayes Classification, for handling data imbalance in emotion classification of tweets. The analysis results show that SMOTE consistently provides a more significant improvement in accuracy, precision, recall, and F1-score compared to RUS, especially in Random Forest and Neural Network models. Maximum Entropy and SVM prove to be the best-performing models in both scenarios, while Naive Bayes Classification, although efficient in terms of time, shows lower performance in evaluation metrics. Overall, SMOTE is a more effective sampling technique compared to RUS in handling class imbalance.