Viewing Model Details

Model details describe model attributes, rules, statistics, and other information about the model. The DBMS_DATA_MINING package supports a separate GET_MODEL_DETAILS function for each algorithm. Global details are also available for Generalized Linear Models, Expectation Maximization, Singular Value Decompostion, and Association Rules.

Model details reverse the transformations applied to the attributes, thus enabling the information to be easily understood by a user. You can obtain the transformations embedded in the model by invoking the DBMS_DATA_MINING.GET_MODEL_TRANSFORMATIONS function.

The query in Example 5-3 returns the coefficients for several attribute values in a GLM regression model called GLMR_SH_Regr_sample. Additional details available for this algorithm include: standard error, test statistic, p value, standard coefficient, lower coefficient limit, and upper coefficient limit.

The query in Example 5-4 returns global details for the same model.

Example 5-3 Model Details for GLM Regression

SELECT attribute_name, attribute_value, coefficient
    FROM TABLE(dbms_data_mining.get_model_details_glm('GLMR_SH_Regr_sample'))
    WHERE attribute_name IN ('AFFINITY_CARD','BULK_PACK_DISKETTES','COUNTRY_NAME')
   ORDER BY class, attribute_name, attribute_value;
 
ATTRIBUTE_NAME       ATTRIBUTE_VALUE      COEFFICIENT
-------------------- -------------------- -----------
AFFINITY_CARD                              -.58234968
BULK_PACK_DISKETTES                        -.99684665
COUNTRY_NAME         Argentina             -1.2032688
COUNTRY_NAME         Australia             .000541598
COUNTRY_NAME         Brazil                5.29534224
COUNTRY_NAME         Canada                4.02414761
COUNTRY_NAME         China                 .878394982
COUNTRY_NAME         Denmark               -2.9852215
COUNTRY_NAME         France                -1.0946872
COUNTRY_NAME         Germany               -1.6345684
COUNTRY_NAME         Italy                 -1.2749328
COUNTRY_NAME         Japan                  -6.259627
COUNTRY_NAME         New Zealand           5.07675762
COUNTRY_NAME         Poland                2.20458524
COUNTRY_NAME         Saudi Arabia          .443146197
COUNTRY_NAME         Singapore             -4.9472244
COUNTRY_NAME         South Africa          .493327068
COUNTRY_NAME         Spain                 -3.0895076
COUNTRY_NAME         Turkey                -5.9014625
COUNTRY_NAME         United Kingdom        2.25154714

Example 5-4 Global Details for GLM Regression

SELECT *
  FROM TABLE(dbms_data_mining.get_model_details_global('GLMR_SH_Regr_sample'))
ORDER BY global_detail_name;

GLOBAL_DETAIL_NAME             GLOBAL_DETAIL_VALUE
------------------------------ -------------------
ADJUSTED_R_SQUARE                             .732
AIC                                       5943.057
COEFF_VAR                                   18.165
CORRECTED_TOTAL_DF                        1499.000
CORRECTED_TOT_SS                        278740.504
DEPENDENT_MEAN                              38.892
ERROR_DF                                  1420.000
ERROR_MEAN_SQUARE                           49.908
ERROR_SUM_SQUARES                        70869.218
F_VALUE                                     52.291
GMSEP                                       52.722
HOCKING_SP                                    .035
J_P                                         52.570
MODEL_CONVERGED                              1.000
MODEL_DF                                    79.000
MODEL_F_P_VALUE                               .000
MODEL_MEAN_SQUARE                         2609.739
MODEL_SUM_SQUARES                       206169.407
NUM_PARAMS                                  80.000
NUM_ROWS                                  1500.000
ROOT_MEAN_SQ                                 7.065
R_SQ                                          .746
SBIC                                      6368.114
VALID_COVARIANCE_MATRIX                       .000