Deep myometrial infiltration, clinical risk score, histological type, and lymphovascular space invasion are important clinical variables that have significant management implications for endometrial cancer patients. Determination of these factors using pure T2-weighted MRI is time-consuming, and the accuracy of this relies on the experience of the clinicians. Combining clinical information and radiomic features from MRI, we developed machine learning classification models to predict these clinical variables. Based on a training dataset, an automatic selection classification model with an optimized hyperparameters method was adopted to find the optimal classifiers. The accuracy of the model predictions was evaluated using an independent external testing dataset. The results suggest that an integrated model (combining clinical and radiomic features) achieved a reasonable accuracy for endometrial cancer clinical variable prediction. The application of these models in clinical practice could potentially lead to cost reductions and personalized treatment.