Aiming at the problems of frequent fault alarm signals and poor system stability of semi-submersible platform, the important factors affecting the stable operation of semi-submersible platform is studied based on the data set of working signal points of semi-submersible platform. Based on these factors, a platform system stability prediction model based on Deep Neural Network (DNN) is constructed by machine learning and deep learning algorithm. The Area Under Curve (AUC) score of the model is improved by 1.0%-16.0% and the accuracy is improved by 3.0%-25.6% compared with those of the traditional machine learning models such as Logistics Regression (LR), K-Nearest Neighbor query (KNN), Support Vector Machine (SVM) and Na?ve Bayesian (Nb), which indicates that the DNN model is of good fitting ability and generalization ability, and can be used in industrial practice.