| 摘要 |
A method for analyzing records of a data base by selecting a target measure related to a selected outcome, identifying data in known records of the data base for use as predictor variables, grouping selected ones of the predictor variables, producing derived values of the target measure for different combinations of the predictor variables for each group, identifying the derived values for a test record, identifying a selected number of known records that are most similar to the test record with respect to the derived values, identifying the value of the selected outcome of the selected most similar known records, and using that value for predicting a selected outcome for the test record. |
