Random forest approach for determining risk prediction and predictive factors of type 2 diabetes: large-scale health check-up data in Japan.

Correct use of the RF method may enable highly accurate risk prediction for the change in HbA1c and may allow the identification of new diabetes risk predictors.