The TREESPLIT Procedure

The "Model Information" table contains the settings of the training parameters. This table also contains some basic information about the resulting decision tree. This table is produced by default.

The "Number of Observations" table contains the number of observations that are read from the input data table and the number of observations that are used in the analysis. When you specify the PARTITION statement, the table also indicates the number of observations that are used in each partition. This table is produced by default.

The "CV Statistics" table contains information about k-fold cross validated cost complexity pruning, such as the alpha values, the corresponding size of the tree, and the corresponding minimum, average, and maximum error. The table is generated when you specify the CVCC option in the PROC TREESPLIT statement, and you are also using k-fold cross validated cost complexity pruning.

The "Tree Performance" table contains statistics that measure the model’s goodness of fit. If you specify the PARTITION statement, the table also indicates these statistics on all partitions of the data.

The "Variable Importance" table displays variable importance based on residual sum of square errors, which is explained in the section Residual Sum of Squares (RSS) Importance Method. This table is produced by default.

The "RBA Variable Importance" table displays variable importance based on the random branch assignment (RBA) method, which is explained in the section Random Branch Assignment Importance Method. This table is produced by the RBAIMP option in the PROC TREESPLIT statement.

The "Reduced Error" table contains the pruning results of reduced error pruning that uses validation data. This is the table of the data shown in the PruningPlot. This table is generated when you specify the PRUNINGTABLE option and are using reduced error pruning in the PRUNE statement with validation data.

The "Cost Complexity" table contains the pruning results of cost complexity pruning that uses validation data. This is the table of the data shown in the PruningPlot. This table is generated when you specify the PRUNINGTABLE option and are using cost complexity pruning in the PRUNE statement with validation data.

When you specify the OUTPUT statement or the OUTMODEL= option in the PROC TREESPLIT statement to create output tables on your CAS server, the TREESPLIT procedure produces the output data table along with a table that lists the CAS library, the data table name, and the number of rows and columns in that data table.

The "Tuner Information" table displays the setup values that the tuner uses. This table is produced by the AUTOTUNE statement.

The "Tuner Summary" table displays statistics about the tuning process. This table is produced by the AUTOTUNE statement.

The "Tuner Timing" table displays the total time spent on different tasks while tuning. This table is produced by the AUTOTUNE statement.

The "Best Configuration" table displays the hyperparameters and objective function values for the best configuration. This table is produced by the AUTOTUNE statement.

The "Tuner Results" table displays the values of the hyperparameters, the objective function for the default configuration (Iteration 0), and up to 10 best found configurations. This table is produced by the AUTOTUNE statement.

The "Evaluation History" tables displays the values of the hyperparameters and the objective function for all configurations. This table is produced by the AUTOTUNE statement, either by default or when EVALHISTORY=ALL.

The "Predicted Probability Names" table indicates the names of the created variables in either the score code or the data table that is specified in the OUT= option in the OUTPUT statement. This table is produced when you specify the PRINTTARGET option in the PROC TREESPLIT statement.