In China, there are many patients with renal failure and the treatment period of renal failure is long, but the number of doctors is not proportional to the number of patients. Renal dialysis is a kind of blood purification technology, which has a certain significance for alleviating the symptoms of patients with renal failure and prolonging the survival period. In this case, it is meaningful to establish an efficient recommendation model to assist the doctor in determining the renal dialysis treatment plan. Based on 22039 renal dialysis treatment records of 180 patients, we proposed a recommendation model of renal dialysis treatment plan based on deep learning, which can recommend a treatment plan based on patient's physical data to provide an intelligent reference for doctors. The model can first intelligently complete the missing data and recommend treatment plans for new patients and re-visiting patients. For new patients, the model based on a deep neural network which mines large data of existing patients, recommend the optimal treatment plan. For re-visiting patients, the model based on the LSTM which personalized learns their personal diagnosis and treatment data, recommend a personalized treatment plan. The model recommends dialyzer mode, dialyzer type, anticoagulant type, and anticoagulant first dose according to the type of disease, complications, dry weight, current weight, age and gender. The model can effectively improve the doctor's diagnosis speed, improve the utilization rate of medical resources and has broad application and promotion prospects.