Support Vector Machine Implementation for Classifying Public Sentiment on Electronic Stamps use in Civil Servant Registration

Civil Servants play a crucial role in government administration and national development. The selection process for Civil Servant Candidates is a critical stage in civil servant management, as it has long-term implications for organizational effectiveness. With the advancement of digital technology, the Indonesian government has implemented electronic stamps in the civil servant candidate registration process to enhance efficiency and transparency. However, this policy has received various responses from the public, particularly on the social media platform Twitter. Many users express positive opinions regarding the convenience of administrative digitalization, while others report technical issues in using electronic stamps. Therefore, sentiment analysis is needed to understand public responses to this policy. One of the effective methods for sentiment classification is Support Vector Machine (SVM), which can optimally separate positive and negative opinions. This study utilizes a dataset comprising 1,249 reviews collected from Twitter/X. The best SVM model is selected through hyperparameter tuning using the GridSearchCV technique. The findings indicate that the SVM model with cost = 100 and gamma = 0.01 achieves the best performance, with an accuracy of 92%, precision of 86.48%, recall of 86.48%, F1-score of 86.48%, and a Kappa-Statistic of 81%.