<![CDATA[Abstrak - Obesitas merupakan salah satu permasalahan kesehatan yang terus meningkat dan berpotensi menimbulkan berbagai penyakit kronis. Pemanfaatan machine learning dapat membantu proses klasifikasi tingkat obesitas berdasarkan karakteristik individu. Penelitian ini bertujuan membandingkan kinerja algoritma Support Vector Machine (SVM) dan Extreme Gradient Boosting (XGBoost) dalam klasifikasi tingkat obesitas serta menganalisis pengaruh Synthetic Minority Oversampling Technique (SMOTE) dan Grid Search terhadap performa kedua algoritma. Dataset yang digunakan adalah Obesity Levels Dataset yang terdiri atas 2.087 data setelah dilakukan penghapusan 24 data duplikat, dengan 16 variabel prediktor dan 7 kelas tingkat obesitas. Tahapan penelitian meliputi preprocessing data, Label Encoding, pembagian data menggunakan Stratified Train-Test Split dengan rasio 80:20, standardisasi menggunakan StandardScaler, penerapan SMOTE pada data latih, optimasi hyperparameter menggunakan Grid Search, serta pelatihan model SVM dan XGBoost. Evaluasi dilakukan menggunakan Accuracy, Precision, Recall, F1-Score, ROC-AUC, Classification Report, dan Confusion Matrix. Hasil penelitian menunjukkan bahwa model XGBoost tanpa SMOTE dan tanpa Grid Search memberikan performa terbaik pada dataset dan skenario eksperimen penelitian ini dengan Accuracy sebesar 97,13%, Precision sebesar 97,15%, Recall sebesar 97,13%, F1-Score sebesar 97,13%, dan ROC-AUC sebesar 99,90%. Pada SVM, penerapan SMOTE meningkatkan Accuracy dari 88,28% menjadi 89,23%, sedangkan kombinasi SMOTE dan Grid Search menghasilkan Accuracy sebesar 96,65%. Sementara itu, penerapan SMOTE dan Grid Search pada XGBoost tidak meningkatkan Accuracy dibandingkan model dasar. Hasil tersebut menunjukkan bahwa efektivitas SMOTE dan Grid Search berbeda pada setiap algoritma. SMOTE dan Grid Search memberikan kontribusi yang lebih besar terhadap peningkatan kinerja SVM, sedangkan XGBoost telah menghasilkan performa yang sangat tinggi pada model dasar. Kata kunci : klasifikasi obesitas; Support Vector Machine; XGBoost; SMOTE; Grid Search; machine learning; Abstract - Obesity is a major health problem that continues to increase and may lead to various chronic diseases. Machine learning can be utilized to support obesity level classification based on individual characteristics. This study aims to compare the performance of Support Vector Machine (SVM) and Extreme Gradient Boosting (XGBoost) algorithms for obesity level classification and to analyze the effects of Synthetic Minority Oversampling Technique (SMOTE) and Grid Search on the performance of both algorithms. The dataset used was the Obesity Levels Dataset consisting of 2,087 records after removing 24 duplicate records, with 16 predictor variables and 7 obesity level classes. The research stages included data preprocessing, Label Encoding, an 80:20 Stratified Train-Test Split, standardization using StandardScaler, SMOTE implementation on the training data, hyperparameter optimization using Grid Search, and model training using SVM and XGBoost. Model performance was evaluated using Accuracy, Precision, Recall, F1-Score, ROC-AUC, Classification Report, and Confusion Matrix. The results show that XGBoost without SMOTE and Grid Search achieved the best performance on the dataset and experimental scenarios used in this study, with an Accuracy of 97.13%, Precision of 97.15%, Recall of 97.13%, F1-Score of 97.13%, and ROC-AUC of 99.90%. For SVM, SMOTE increased Accuracy from 88.28% to 89.23%, while the combination of SMOTE and Grid Search achieved an Accuracy of 96.65%. Meanwhile, the application of SMOTE and Grid Search to XGBoost did not improve Accuracy compared with the baseline model. These results indicate that the effectiveness of SMOTE and Grid Search differs between algorithms. SMOTE and Grid Search provided a greater contribution to improving SVM performance, whereas XGBoost already achieved very high performance using its baseline model. Keywords: obesity classification; Support Vector Machine; XGBoost; SMOTE; Grid Search; machine learning;]]>
