<![CDATA[Purpose: This study aims to analyze user sentiment regarding the Threads application by comparing the performance of different machine learning models. As a relatively new social media platform, understanding user feedback is crucial for identifying service gaps and improving user retention. The research seeks to determine which algorithm provides the highest precision in classifying user reviews into positive and negative sentiments. Methods: The research utilized a dataset of 3,000 user reviews scraped fromthe Google Play Store. The methodology followed a systematic text mining workflow, including preprocessing stages such as noise removal, tokenization, stopword removal, and stemming. Feature extraction was performed using the Term Frequency-Inverse Document Frequency (TF IDF) method. Three machine learning algorithms—Support Vector Machine (SVM), Decision Tree, and Logistic Regression—were implemented and evaluated using K-Fold Cross Validation to ensure statistical reliability. Result: The experimental results indicate that the Support Vector Machine (SVM) consistently outperformed the other two models. SVM achieved a superior average accuracy of 88.18%, with a peak performance reaching 92.69% during K-Fold testing. Logistic Regression and Decision Tree showed lower accuracy and less stability in handling the high-dimensional text data. These figures confirm that SVM is the most effective model for analyzing the linguistic nuances found in Threads app reviews. Novelty/Originality/Value: This research contributes to the field of software evaluation by providing an empirical comparison of classification algorithms specifically for newly launched social media platforms like Threads. The findings offer practical value for developers to automate the monitoring of user satisfaction. The study demonstrates that integrating rigorous TF-IDF weighting with SVM significantly enhances the accuracy of sentiment detection in short-form mobile application reviews.]]>
Copyrights © 2025
