SAS Enterprise Miner

User : Dan
Date : 28JUL2006:18:43:52
Notes:

"EM Workspace" :

EM Workspace

SASUSER.SMALLD

Input Data Settings:

All variables

Interval Variables

Class Variables

Notes: not available

Data Partition

Partition Settings

Output

Log

Training Code

Notes

Transform Variables

Interval Variables and Transformations

Notes

NN-D3

Optimization plot:

Optimization

 
 Fit Statistic                     Training    Validation       Test 
  
 [ TARGET=RESP_FLG ]                    .            .           . 
 Average Profit                        0.28         0.28        0.30 
 Misclassification Rate                0.28         0.28        0.30 
 Average Error                         0.57         0.57        0.58 
 Average Squared Error                 0.19         0.19        0.20 
 Sum of Squared Errors              1375.19      1223.42      472.68 
 Root Average Squared Error            0.44         0.44        0.44 
 Root Mean Squared Error                .           0.44        0.44 
 Error Function                     4076.87      3623.09     1389.99 
 Mean Squared Error                     .           0.19        0.20 
 Maximum Absolute Error                0.88         0.87        0.87 
 Divisor for ASE                    7200.00      6400.00     2400.00 
 Model Degrees of Freedom           6391.00          .           . 
 Degrees of Freedom for Error      -2791.00          .           . 
 Total Degrees of Freedom           3600.00          .           . 
 Sum of Frequencies                 3600.00      3200.00     1200.00 
 Sum Case Weights * Frequencies     7200.00      6400.00     2400.00

Network settings

Objective function: Maximize profit
Assessment Matrix: Profit vector
Utilities

Variables

Output

Log

Training Code

Score Code

Model assessment settings
Train data set is not selected for assessment.
Validation data set is selected for assessment.
Test data set is not selected for assessment.
Scored data set: 5000 observations are saved for interactive model assessment.

SAS Graphics

Confusion Matrix (Assessed Partition=VALIDATION)

Notes

NN-D2

Optimization plot:

Optimization

 
 Fit Statistic                     Training    Validation       Test 
  
 [ TARGET=RESP_FLG ]                    .            .           . 
 Average Profit                        0.28         0.28        0.30 
 Misclassification Rate                0.28         0.28        0.30 
 Average Error                         0.59         0.59        0.61 
 Average Squared Error                 0.20         0.20        0.21 
 Sum of Squared Errors              1441.78      1290.22      501.68 
 Root Average Squared Error            0.45         0.45        0.46 
 Root Final Prediction Error           1.83          .           . 
 Root Mean Squared Error               1.33         0.45        0.46 
 Error Function                     4248.34      3794.85     1461.46 
 Mean Squared Error                    1.77         0.20        0.21 
 Maximum Absolute Error                0.72         0.72        0.72 
 Final Prediction Error                3.34          .           . 
 Divisor for ASE                    7200.00      6400.00     2400.00 
 Model Degrees of Freedom           3193.00          .           . 
 Degrees of Freedom for Error        407.00          .           . 
 Total Degrees of Freedom           3600.00          .           . 
 Sum of Frequencies                 3600.00      3200.00     1200.00 
 Sum Case Weights * Frequencies     7200.00      6400.00     2400.00 
 Akaike's Information Criterion    10634.34          .           . 
 Schwarz's Baysian Criterion       30394.82          .           .

Network settings

Objective function: Maximize profit
Assessment Matrix: Profit vector
Utilities

Variables

Output

Log

Training Code

SAS Graphics

Confusion Matrix (Assessed Partition=VALIDATION)

Notes

Assessment-NN

SAS Graphics

Variable Selection

Results

Settings:

Manual selections

Target associations

Missing values

Hierarchies

Variables

Output

Log

Training Code

Score Code

Notes

Logit-D2

Parameters:

Estimates Table

Fit Statistics

 
 Fit Statistic                         Training      Validation            Test 
  
 Akaike's Information Criterion    4000.3154453               .               . 
 Average Squared Error             0.1831276886    0.1926826553    0.1996994379 
 Average Error Function            0.5464327007    0.5690393158    0.5855879534 
 Degrees of Freedom for Error              3567               .               . 
 Model Degrees of Freedom                    33               .               . 
 Total Degrees of Freedom                  3600               .               . 
 Divisor for ASE                           7200            6400            2400 
 Error Function                    3934.3154453    3641.8516211    1405.4110881 
 Final Prediction Error            0.1865160899               .               . 
 Maximum Absolute Error            0.9690954513    0.9541993297    0.9340076785 
 Mean Square Error                 0.1848218892    0.1926826553    0.1996994379 
 Sum of Frequencies                        3600            3200            1200 
 Number of Estimate Weights                  33               .               . 
 Root Average Sum of Squares       0.4279342106    0.4389563251    0.4468774306 
 Root Final Prediction Error        0.431875086               .               . 
 Root Mean Squared Error           0.4299091639    0.4389563251    0.4468774306 
 Schwarz's Bayesian Criterion      4204.5421864               .               . 
 Sum of Squared Errors             1318.5193577     1233.168994    479.27865108 
 Sum of Case Weights Times Freq            7200            6400            2400 
 Misclassification Rate            0.2672222222       0.2821875             0.3 
 Total Profit for RESP_FLG                  997             896             356 
 Average Profit for RESP_FLG       0.2769444444            0.28    0.2966666667

Target Information:

Objective function: Maximize profit
Assessment Matrix: Profit vector
Utilities

Regression Settings:

Output

Log

Training Code

Score Code
Model assessment settings
Train data set is not selected for assessment.
Validation data set is selected for assessment.
Test data set is not selected for assessment.
Scored data set: 5000 observations are saved for interactive model assessment.

SAS Graphics

Confusion Matrix (Assessed Partition=VALIDATION)

Notes

Tree-D2

Model assessment plot:

SAS Graphics

 
 Fit Statistic                     Training    Validation       Test 
  
 Average Squared Error                 0.18         0.19        0.21 
 Sum of Squared Errors              1316.15      1213.74      495.03 
 Root Average Squared Error            0.43         0.44        0.45 
 Maximum Absolute Error                0.94         0.94        0.94 
 Divisor for ASE                    7200.00      6400.00     2400.00 
 Total Degrees of Freedom           3600.00          .           . 
 Misclassification Rate                0.27         0.28        0.29 
 Number of Estimated Weights           9.00          .           . 
 Sum of Frequencies                 3600.00      3200.00     1200.00 
 Sum Case Weights * Frequencies     7200.00      6400.00     2400.00

 
                          N *             V N * 
 Node    Leaf     N     PRIORS    V N    PRIORS     % V 0     % V 1       % 0       % 1 
  
   4       1     157      157     149      149      45.64     54.36     43.95     56.05 
  10       2     431      431     351      351      70.94     29.06     73.78     26.22 
  11       3     699      699     614      614      61.56     38.44     58.66     41.34 
   6       4     308      308     326      326      64.11     35.89     59.42     40.58 
  50       5      47       47      51       51      60.78     39.22     61.70     38.30 
  82       6     974      974     911      911      82.22     17.78     84.80     15.20 
  83       7     125      125      93       93      70.97     29.03     71.20     28.80 
  27       8     340      340     276      276      92.03      7.97     93.82      6.18 
  15       9     519      519     429      429      69.93     30.07     69.36     30.64

English rules

Sequence

Matrix

Target information
Name: RESP_FLG
Label: Responder Flag(1:responder; 0:nonrespond
Measurement: binary

Tree settings

Objective function: Maximize profit
Assessment Matrix: Profit vector
Utilities

Splitting criterion: Gini Reduction
Minimum number of observations in a leaf: 15
Observations required for a split search: 31
Maximum number of branches from a node: 2
Maximum depth of tree: 6
Splitting rules saved in each node: 5
Surrogate rules saved in each node: 0
Treat missing as an acceptable value
Model assessment measure: Average Square Error (Gini index)
Subtree: Best assessment value
Observations sufficient for split search: 3600
Maximum tries in an exhaustive split search: 5000
Do not use profit matrix during split search
Do not use prior probability in split search

Log

Score Code
Model assessment settings
Train data set is not selected for assessment.
Validation data set is selected for assessment.
Test data set is not selected for assessment.
Scored data set: 5000 observations are saved for interactive model assessment.

SAS Graphics

Confusion Matrix (Assessed Partition=VALIDATION)

Notes

Ensemble [Combined]

Settings

 
  Model_type             = COMBINED 
  Number of models       = 3 
  Number of loops        = 1 
  Target variable name   = RESP_FLG 
  Target variable level  = binary 
  Target variable type   = num 
  Target variable format = BEST12. 
  Frequency variable     = 
  From variable          = F_RESP_FLG 
  Into variable          = I_RESP_FLG 
  
  Input data             : Training   = EMDATA.STRN3HZA 
                           Validation = EMDATA.SVAL7TRX 
                           Testing    = EMDATA.STST2J13 
  
  Output data            : Training   = EMDATA.STRN1HC4 
                           Validation = EMDATA.SVAL7Z82 
                           Testing    = EMDATA.STSTACK9 
  
  Process train data     = No 
  Process score data     = Yes

Variables

Output

Log

Training Code

SAS Graphics