<![CDATA[This study aims to analyze review sentiment and compare the performance of the Decision Tree and Support Vector Machine (SVM) algorithms on a dataset of 14,000 reviews from the m.tix – XXI app, which were classified as 57.4% positive, 33.8% negative, and 8.8% neutral. Through the pre-processing stage, 200 data points were deemed invalid, leaving 13,800 data points suitable for analysis. The dataset was then split into two categories: 80% training data (11,040 reviews) and 20% testing data (2,760 reviews), to support more accurate model performance measurement. The main contribution of this study lies in identifying the advantages of SVM in handling review data, with evaluation results showing that SVM achieved an accuracy of 86%. This is evidenced by the significant superiority of SVM’s F1-score across all categories, particularly for positive sentiment (0.93), negative sentiment (0.80), and neutral sentiment (0.08) compared to the Decision Tree’s accuracy of 83%, with F1-scores of 0.92 for positive sentiment, 0.73 for negative sentiment, and 0.03 for neutral sentiment. This research can be utilized by m.tix-XXI management as a foundation for evaluating and improving the quality of the m.tix-XXI application’s services.]]>
Copyrights © 2026
