The existing gastric cancer (GC) risk prediction models based on biomarkers are limited. This study aims to identify new promising biomarkers for GC to develop a risk prediction model for effective assessment, screening, and early diagnosis. This study was conducted utilizing a large combined cohort for upper gastrointestinal cancer that was established in Hebei Province, China. General macro risk factors, Helicobacter pylori (H.pylori) infection status, and protein biomarkers were collected through questionnaire surveys and laboratory tests. Novel GC biomarkers were explored using data-independent acquisition (DIA) proteomics and enzyme-linked immunosorbent assay (ELISA). Multiple machine learning algorithms were used to identify key predictors for the GC risk prediction model, which was validated with an independent external cohort from multiple hospitals. A total of 530 participants aged 40 to 74 were analyzed, with 104 ultimately diagnosed with GC. Significant biomarkers in GC patients were identified by DIA combined ELISA, including elevated Keratin 7 (KRT7) and Mammary fibrostatin (SERPINB5) (P<0.001) and decreased Dickkopf-associated protein 3 (DKK3) (P<0.001). Factors such as sex, age, smoking status, alcohol consumption, family history of GC, H. pylori infection, DKK3 and SERPINB5 were used to create a multidimensional risk prediction model for GC. This model achieved an area under the curve (AUC) of 0.938 (95% confidence interval: 0.913-0.962). The risk prediction model developed in this study shows high accuracy and practical utility, serving as an effective preliminary screening tool for identifying high-risk individuals for GC.