Objective To set up a diagnostic model of gastric cancer by using biomarkers for differential diagnosis of early gastric cancer through statistics and bioinformatics technologies. Methods We used retrospective cohort study to screening 179 cases of gastric associated diseases.139 cases of gastric cancer and 40 cases of atypical hyperplasia were separated by the diagnosis of gastroscopy. The serological parameter differences were analyzed by statistical multi-parameter analysis between gastric cancer and atypical hyperplasia. SPSS18.0 software was adopted for analysis of tumor markers cytokines and biochemical indicators in two groups to set up a diagnostic model. Results There were significant difference in 20 parameters between gastric cancer and atypical hyperplasia (P < 0. 05). The diagnosis accuracy of single marker was low. The diagnosis horizontal of the IL-6 and CEA were highest. The AUC was 0.824 and 0.721, respectively .The results of logistic regression analysis showed CEA, CA72-4, IL-6, IMA, SOD and ApoA1 had significant differences in the differential diagnosis for gastric cancer, and the prediction accuracy of two groups was 85.3%. Conclusion This study provides a promising strategy for differential diagnosis of gastric cancer with statistics software. The parameters of CEA，CA72-4，IMA，SOD and ApoA1 may have important roles in the predicted classification of gastric cancer and atypical hyperplasia with linear pattern.