Sentiment Analysis of Imbalanced Tourism News Dataset Using Random Forest with Particle Swarm Optimization and Synthetic Minority Oversampling
Husni Husni, Arif Muntasa, Vina Angelina Savitri · 2024
This paper reports the results of research on the classification of Indonesian tourism news texts based on their sentiment using the Random Forest method combined with feature selection and sampling techniques. The dataset comprised lots of tourism news related to Madura, East Java. Feature selection in the classification process was performed using Particle Swarm Optimization (PSO) to identify influential features. Subsequently, the dataset was divided into two parts: training and testing data. In the training data, an imbalance in the number of positive and negative classes led to classification results biased towards the majority class. Therefore, this study employed the Synthetic Minority Oversampling Technique (SMOTE) to balance class numbers in the training data. Following that, the classification process utilized the Random Forest method to determine the accuracy of this study. The obtained results revealed an accuracy rate of 91% as the average accuracy when using the PSO and Random Forest methods. The PSO feature selection method contributed to accelerating computation time compared to not using feature selection methods.