ICU (Intensive Care Unit) provides isolation places and equipment for severely ill or unconscious patients, and provides services such as the best nursing, comprehensive treatment, and early postoperative rehabilitation. A large amount of data is generated every day in the ICU, including demographic information, admission examination, diagnosis, medication and other information. We establish a prediction model for NSICU (Neurosurgical Intensive Care Unit) patient using machine learning. In this paper, we choose five classical model to determine the mortality prediction model of NSICU patients. In order to alleviate the problem of class imbalance, this article resamples the sample to eliminate the impact of this problem based on SMOTEENN. In addition, use the SHapley Additive explanation (SHAP) interpreter to illustrate the relationship between predictive variables and the results, which is helpful for doctors to make decisions.