A data mining technique that analyzes existing data and uses the results to predict values of attributes for new records or missing attributes in existing records. For example, existing credit application data can be used to predict the credit risk for a new application.