OBJECTIVES: In this study, we used artificial neural network---a popular data mining tool, to build a predictive model for risk of diabetic retinopathy. We compared its performance with Logistic regression in terms of their discrimination capacity. METHODS: National Health and Nutrition Examination Survey (NHANES) data was used in this study. Participants with a diabetes diagnosis and a known retinopathy status (yes/no) were included. Two models were built using training sample: artificial neural network and logistic regression. We used these two models to predict the risk of diabetic retinopathy in the testing sample. Receiver operating characteristic (ROC) were calculated and compared for these two models for their discrimination capability. RESULTS: A total of 757 patients were recruited and 21.5 % had retinopathy. A random sample of 400 was chosen as the testing sample and the rest was used as the training sample. The Area Under the Curve (AUC) is about 0.75 for training sample according to above logistic regression, meaning that a randomly selected individual from the positive group has a test value larger than that for a randomly chosen individual from the negative group 75 percent of the time. After logistic regression and network analysis were conducted in the training sample, we used the outputs from both models to predict the likelihood in the testing sample (N=400). The areas under the receiver operating characteristic curves were 0.72 and 0.73 for the logistic model and the neural network, respectively. There were no significant differences in predictive ability between the approaches. CONCLUSIONS: This study suggests that it is possible to develop a reproducible and transportable predictive instrument for diabetes patients with retinopathy complication. In our research, both logistic regression and neural network models did a good job of predicting the risk for retinopathy complication.