This study aimed to revalidate determinants influencing the restoration of tongue pressure in older individuals following therapy for head and neck malignancies, utilizing advanced machine learning approaches. Logistic regression, support vector regression, random forest, and extreme gradient boosting models were trained on variables such as age, surgical category, dental condition, and demographic characteristics, derived from detailed patient records and direct tongue pressure measurements. Results: Logistic regression provided the highest predictive accuracy, yielding an accuracy of 0.630 [95% CI: 0.370–0.778], an F1 score of 0.688 [95% CI: 0.435–0.853], a precision of 0.611 [95% CI: 0.313–0.801], a recall of 0.786 [95% CI: 0.413–0.938], and an AUC of 0.626 [95% CI: 0.409–0.806]. The most significant predictors included glossectomy (p = 0.039), the number of functional teeth (p = 0.043), and patient age (p = 0.044), with significance set at p < 0.05. The findings confirmed that glossectomy, functional dentition, and age were key variables influencing tongue pressure in logistic regression, while the presence of natural teeth and tumors situated on the tongue remained consistent predictors across all algorithms assessed.
