Abstract: In order to reduce the dimensionality of the dataset to obtain the best feature subset as well as to improve the accuracy of the prognostic survival classification of gastric cancer， a hybrid network model of deep belief network and support vector machine combined with feature selection algorithm was proposed. Based on the filtered feature selection algorithm， a distance coefficient was introduced to adjust the overall degree of bias and reduce the instability of the calculated weight values， so as to construct new sample weight values， and then analyze the subset of features that have a greater impact on the survival period of gastric cancer through the Pearson’s correlation coefficient； The constrained Boltzmann machine module was adopted in the deep belief network， and then the subset of features in the hidden layer was subjected to the feature extraction； Finally， the support vector machine was used to classify the output values of the last layer of the deep belief network to realize the classification of gastric cancer survival. By improving the feature selection algorithm and combining the advantages of deep belief network and support vector machine， the model showed better accuracy， AUC value and F1 value in the experiments， which are 81.2%， 83.4% and 81.5%， respectively， compared with the traditional single machine learning method.