<![CDATA[Abstrak : Tenaga pendidik memiliki peran strategis dalam kemajuan pendidikan nasional, sehingga pemahaman terhadap sentimen mereka penting dalam meningkatkan kesejahteraan dan kualitas layanan pendidikan. Dalam pengolahan data sentimen, seringkali ditemui tantangan seperti ketidakseimbangan data antara sentimen mayoritas dan minoritas, serta tingginya jumlah fitur yang menyebabkan dimensionalitas data menjadi besar.Tujuan penelitian ini adalah menganalisis sentimen attitude tenaga pendidik di Indonesia menggunakan metode klasifikasi Machine Learning, dengan pendekatan seleksi fitur berbasis korelasi dan penyeimbangan data melalui Synthetic Minority Over-sampling Technique–Edited Nearest Neighbours (SMOTE ENN). Model klasifikasi dibangun menggunakan algoritma Naïve Bayes dan Support Vector Machine (SVM).Hasil penelitian menunjukkan bahwa SVM memberikan akurasi lebih tinggi dibandingkan Naïve Bayes, baik pada data asli maupun setelah penerapan SMOTE ENN. Akurasi Naïve Bayes meningkat dari 61% menjadi 89% setelah seleksi fitur berbasis korelasi, sedangkan SVM meningkat dari 69% menjadi 97%. Penelitian ini membuktikan bahwa kombinasi SVM, SMOTE ENN, dan seleksi fitur berbasis korelasi mampu meningkatkan akurasi klasifikasi sentimen tenaga pendidik di Indonesia secara signifikan. Kata kunci : analisis sentimen, tenaga pendidik, SVM, SMOTE, seleksi fitur Abstract : Educators play a strategic role in the progress of national education, so understanding their sentiments is important for improving their well-being and the quality of educational services. In sentiment data processing, challenges are often encountered, such as data imbalance between majority and minority sentiments, and a high number of features leading to high data dimensionality. The purpose of this study is to analyze the sentiment of Indonesian educators' attitudes using Machine Learning classification methods, with a correlation-based feature selection approach and data balancing through Synthetic Minority Over-sampling Technique–Edited Nearest Neighbours (SMOTE ENN). Classification models were built using the Naïve Bayes and Support Vector Machine (SVM) algorithms. The research results show that SVM provides higher accuracy compared to Naïve Bayes, both on the original data and after applying SMOTE ENN. Naïve Bayes' accuracy increased from 61% to 89% after correlation-based feature selection, while SVM's increased from 69% to 97%. This study proves that the combination of SVM, SMOTE ENN, and correlation-based feature selection can significantly improve the accuracy of sentiment classification for Indonesian educators Keywords: Sentiment analysis, Educators, SMOTE, Feature selection]]>
