Analisis Sentimen Pengguna Media Sosial X terhadap Layanan Pesan Antar Makanan Menggunakan Machine Learning

Food delivery services have become an integral part of modern lifestyle, requiring companies to continuously monitor customer satisfaction thrugh social media X as a platform for real-time opinions. This research aims to analyze user setiment and cimpare the performance of Naive Bayes, Support vector machine (SVM), adan a Deep Learning model (IndoBERT) in classifying these sentiments. The initial dataset consisted of 11,124 tweets collected from July 2023 to July 024, which were filtered into 3,139 clean dara points. Three experimental scenarios were impllemented, including the application of ContextualWordEmbsAug before the data splitting stage to address the imbalanced dataset in Scenario 3. Results show that negative sentiment dominates public perception at 51.5%, followed by positive (26.8%) and neutral (21.7%) sentiments. Model evaluation indicates that the SVM algorithm in Scenario 3 is the most effective, achieving an Accuracy by 11% compared to the baseline and enhanced the model’s robustness in recoginizing minoriy classes. This study provides a strategic baseline regarding post-pandemic consumer liyalty patterns, which remains highly relevant for current service evaluation.