<![CDATA[The rapid development of artificial intelligence (AI) applications has increased the demand for sentiment analysis of user reviews, particularly on the Google Play Store. This study aims to compare the performance of four classical machine learning algorithms—Logistic Regression, Support Vector Machine (SVM), Naïve Bayes, and K-Nearest Neighbors (KNN)—in classifying user sentiments toward the Grok AI application. A total of 2,426 reviews were collected through Google Play Store scraping and processed using several preprocessing steps, including case folding, cleaning, tokenization, stopword removal, normalization, and stemming. Sentiment labels were assigned based on user ratings, while feature representation was conducted using Term Frequency–Inverse Document Frequency (TF-IDF). To address class imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) was applied. Model evaluation employed accuracy, precision, recall, F1-score, and confusion matrix. The experimental results show that SVM achieved the best performance with 88% accuracy, 0.87 precision, 0.86 recall, and 0.86 F1-score. Logistic Regression ranked second with 86% accuracy, followed by Naïve Bayes (81%) and KNN (78%). These findings indicate that SVM is the most effective algorithm for sentiment analysis of AI-based application reviews, while Logistic Regression provides a stable and interpretable alternative. This research contributes by providing a benchmark for the performance of classical machine learning algorithms in the context of Grok AI reviews and offers methodological insights for developers to enhance the quality of AI-based applications.]]>
Copyrights © 2025
