Analisis Sentimen Publik Program PPPK di Media Sosial X menggunakan Naïve Bayes dan SVM

Abstract

Sentiment analysis of the Government Employee Program with Work Agreement (PPPK) is important to understand public perception and as a basis for policy evaluation. This study aims to analyze public sentiment towards the PPPK policy and evaluate the performance of the Support Vector Machine (SVM) and Naïve Bayes algorithms in classifying public opinion on social media X. This study is a quantitative study with a data mining approach. The stages begin with collecting data collection of 7,508 tweets and processed through the stages of preprocessing, labeling, feature extraction using TF-IDF, and classification with SVM and Naïve Bayes. Data balancing is done using the Synthetic Minority Oversampling Technique (SMOTE). Our findings show that SVM produces the highest accuracy of 95%, while Naïve Bayes reaches 87%. The application of SMOTE has been shown to improve the performance of both models, especially in recognizing negative sentiment. The advantage of SVM lies in its ability to optimally separate classes through maximum margin, which is effective for high-dimensional text data. Meanwhile, SMOTE plays an important role in balancing class distribution, thereby increasing accuracy, precision, and recall. These findings provide an important basis for policy makers to respond to public opinion more appropriately based on valid and representative data.