Objective To develop and validate a risk prediction model for radiotherapy interruption in head and neck cancer (HNC) patients based on multidimensional nutritional features, and to provide a reference for identifying the patients at high risk. Methods A total of 412 HNC patients undergoing radiotherapy were enrolled.Five machine learning algorithms-Lasso regression, random forest, support vector machine, extreme gradient boosting, and K-nearest neighbors (KNN)-were employed to construct prediction models for radiotherapy interruption.The predictive performance of these models was compared, and the SHapley Additive exPlanations (SHAP) method was performed to interpret the feature contributions of predictors in the optimal model. Results Univariate analysis revealed that skeletal muscle mass index (SMI), NRS2002 score (nutritional risk screening), prognostic nutritional index (PNI), tumor site, and age were significantly associated with radiotherapy interruption.Among the five algorithms, the KNN model demonstrated the best performance:the area under the receiver operating characteristic curve (AUC) was 0.943 (95% CI:0.926~0.961) in the training set and 0.823 (95%CI:0.734~0.912) in the validation set.SHAP analysis indicated that low SMI and an NRS2002 score ≥3 were the most critical predictors. Conclusion The KNN model developed in this study reliably predicts the risk of radiotherapy interruption in HNC patients and highlights the vital importance of nutrition-related factors in the prediction.