Abstract: To address the need for metabolic disorder assessment, an evaluation model of metabolic flexibility is developed based on the dynamic characteristics of the respiratory quotient (RQ). Twenty metabolically flexible and thirty-six metabolically inflexible participants are recruited from the First Affiliated Hospital of the University of Science and Technology of China. Clinical data are collected. Indirect calorimetry is used to measure RQ values at five time points during an oral glucose tolerance test. Features derived from RQ are extracted. A staged feature selection process is conducted to identify the most relevant predictors. The synthetic minority over-sampling technique (SMOTE) is applied to alleviate class imbalance. Evaluation models are established using logistic regression (LR), k-nearest neighbors (KNN), XGBoost, and random forest (RF) algorithms. All models demonstrate strong performance, with comprehensive scores exceeding 0.920. The RF model achieves the best performance, with an accuracy of 0.941, an F1-score of 0.957, an AUC of 0.985, and a comprehensive score of 0.961. The proposed method combines high predictive accuracy with physiological interpretability, providing a reliable basis for early screening of metabolic disease risk.