This study aimed to provide the epidemiological model evaluating the risk of developing type 2 diabetes (T2DM) in Taiwan periodic health-check population. We derived risk functions using multivariate Cox regression in a random half of the sample. Rules based on these risk functions were evaluated in another half. Model coefficients were used to assign each variable a score. 73,961 subjects aged 35-74, were included and followed up with a median 3.15 years. Six predictive models (PMs) were developed. PM1 contained simple clinical information, while PM2 contained fasting plasma glucose (FPG) based on PM1, and PM3 further added variables indicating lipid level, liver and kidney. PM4 only included FPG. The capability of published ARIC score model was also evaluated. Eventually we considered score defined nine predictors by PM2. The area under the ROC curve (AUC) was 0.848 (95% CI, 0.829-0.868) predicting diabetes within 5 years, and also had adequate performance in validation subsample (AUC=0.833, 95% CI, 0.811-0.855). The 5-year T2DM probability can be calculated by: 1-0.9743960037 exp((score points -15.0281284)). We concluded that this diabetes risk score, derived from clinical information combined with FPG is a simple, effective tool to identify individuals at high risk for undiagnosed T2DM.