Objectives: The urea reduction ratio (URR) is a common measure for determining hemodialysis adequacy. Two blood samples are required, one before dialysis begins and one after dialysis is completed, and the URR is not known until laboratory tests are completed. As a result, determining whether treatment is adequate for each session is challenging. This study developed a neural network model to predict URR based on minimal data from hemodialysis record utilized in clinical practice. Methods: Fifty-two patients with end-stage renal disease were given data from their two hundred forty eight hemodialysis records for fifteen months, as well as URR values measured from blood samples. Machine learning methods including multivariate linear regression, random forest, extreme gradient boosting, and neural network was performed to predict URR. The demographics of patients, and also the data obtained for each hemodialysis treatment, were used to train machine learning models. After training, the performance of each model was evaluated by Spearman's correlation, root mean square error, and mean absolute error using stratified test data. Results: The best performance was achieved by the fully connected neural network model (MAE = 2.413; RMSE = 2.958; Corr = 0.846). Ensemble approaches such as extreme gradient boosting(MAE = 2.528; RMSE = 3.026; Corr = 0.813) and random forest(MAE = 2.443; RMSE = 3.051; Corr = 0.792) produced comparable results. Multivariate linear regression(MAE = 2.504; RMSE = 3.277; Corr = 0.716) showed the lowest performance. Conclusions: Using ordinary data from hemodialysis records, the neural network model can effectively and easily estimate the adequacy of treatment at the patient's bedside.