The nude mice injected with human gastric cancer cells (SGC-7901) in their peritoneums are chosen as animal models of gastric cancer peritoneal dissemination in this research. The Raman spectra at 785 nm excitation of both these nude mice which are in different tumor planting periods and the normal counterpart are taken in vivo in the imitate laparotomy. 205 spectra are collected. The spectra of different tissue types are compared and classified by support vector machine (SVM) algorithm. The results show significant differences between normal and malignant tissues. For normal and malignant tissues, the sensitivity, specificity and accuracy are 95.73%, 70.73% and 90.73%, respectively, while for different tumor planting periods, they are 98.82%, 98.73% and 98.78%. The experimental results show that Raman spectra differ significantly between cancerous and normal gastric tissues, which provides experimental basis for the diagnosis of gastric cancer by Raman spectroscopy technology. And SVM algorithm can give well generalized classification performance for the samples, which expands the application of mathematical algorithms in classification.