Skripsi
ANALISIS SENTIMEN ULASAN PENGGUNA PADA APLIKASI M-PAJAK MENGGUNAKAN METODE RANDOM FOREST DAN XGBOOST
This study aims to evaluate the quality of digital taxation services by analyzing the sentiments contained in M-Pajak application user reviews. Utilizing a dataset of 6,829 reviews classified into negative, neutral, and positive categories, this research examines the performance of the Random Forest and XGBoost algorithms. Although the test results indicate high accuracy levels, reaching 85.21% in the hold-out validation scheme and 90.14% in the stratified k-fold cross-validation, an in-depth evaluation through a confusion matrix reveals a significant model bias toward the majority class. A key finding of this research suggests that the accuracy values are misleading, as both models failed completely to classify the neutral class, yielding very low F1-scores (0.00–0.11). This phenomenon confirms that the primary issue lies not in the choice of algorithm, but in the extreme data distribution imbalance and the ambiguity of rating-based labeling. The scientific contribution of this research lies in proving that the evaluation of sentiment classification systems must go beyond conventional accuracy metrics. By prioritizing performance stability across each class, the resulting system is expected to provide a fairer and more objective evaluation of public data.
No other version available