<![CDATA[The self-photo (photobox) business in Indonesia has grown rapidly as a digital-era experience service, propelled largely by user-generated content (UGC) that acts as electronic word-of-mouth (eWOM) on TikTok. For small businesses, the public opinion embedded in these comments is a valuable yet underused marketing asset. This study analyses consumer sentiment toward photobox services and positions the analysis as a social-media-analytics tool to support digital marketing decisions. A machine-learning approach was applied to 1.086 labelled Indonesian TikTok comments grouped into positive, neutral, and negative classes. Comments were represented with TF-IDF features and classified using Naive Bayes, Logistic Regression, Random Forest, and Support Vector Machine, evaluated through five-fold stratified cross-validation, with SMOTE and SMOTE with Tomek Links applied to handle class imbalance. The results show that the eWOM is overwhelmingly positive (64,0%) and that most comments are extremely short. The best accuracy reached about 63,2%, but every model struggled with the neutral and negative classes; Logistic Regression with SMOTE detected negative opinion best. Read through the Elaboration Likelihood Model, the brevity of the comments reflects low-effort, peripheral-route reactions common on short-video platforms. Managerially, the strong positive buzz signals healthy brand momentum, yet the scarcity of substantive feedback means business owners must combine sentiment analytics with other listening channels to obtain actionable insight.]]>
Copyrights © 2026
