<![CDATA[User reviews on the Steam platform contain important information about player perceptions of EA FC 26. However, the large volume of reviews and their unstructured textual form make manual analysis difficult and prone to inconsistency. This study aims to identify user sentiment tendencies and compare the performance of Support Vector Machine (SVM) and Multinomial Naïve Bayes in classifying EA FC 26 reviews on Steam. The research follows the Knowledge Discovery in Database (KDD) stages, consisting of selection, preprocessing, transformation, data mining, and evaluation. A total of 5,000 recent English-language reviews were collected through the Steam Review API. After data processing, 4,988 valid reviews were obtained, consisting of 2,823 negative reviews and 2,165 positive reviews based on the voted_up recommendation status. The review texts were represented using TF-IDF N-Gram features consisting of unigrams and bigrams, then classified using SVM and Multinomial Naïve Bayes. The evaluation results show that SVM achieved an accuracy of 82.26% and a weighted F1-score of 0.8230, while Multinomial Naïve Bayes achieved an accuracy of 77.15% and a weighted F1-score of 0.7598. The McNemar test produced a p-value of 0.000421, indicating a statistically significant difference between the two models. This study contributes by providing a comparative evaluation based not only on general accuracy, but also on per-class performance, confusion matrix, prediction error patterns, and statistical significance.]]>
Copyrights © 2026
