利用Tree Node合并名义变量及筛选变量

1. Collapsing Levels
   1. Change the maximum number of branches from a node from 2 to the number of levels of variable.
   2. Set the maximum depth of tree to 1.
   3. Set significance level to 1.That is to say, use a large P-value.
   4. Deactivate the Kass adjustment.

   5. Change the sub-tree option to the most leaves to prevent pruning.
   6. Change Model assessment measure from Automatic to Average profit.

   

2. Selecting important variable

   Importance is a weighted average of the reduction in Gini impurity for the surrogate splits using the jth input across all the internal nodes in the tree. The weights are the node sizes.

   

   1. Select the Basic tab,select Gini reduction for the splitting criterion.
   2. Change the maximum depth of the tree to 8 to allow a larger tree to grow.

      

   3. Select the Advanced tab and change the model assessment measure to Total Leaf Impurity(Gini index).
   4. Change the maximum depth of the tree to 8 to allow a larger tree to grow.
   5. Select the Advanced tab and change the model assessment measure to Total Leaf Impurity(Gini index).

      

   6. Select the Score tab and then the Variables subtab.

      

   7. Close the tree nide,saving changes when prompted.

转载于:https://www.cnblogs.com/zgw21cn/archive/2009/04/03/1428905.html