Practical Predictive Analytics

Producing a larger tree

Sometimes with small trees, the decision rules can be too trivial or obvious. In this can, it often makes sense to increase the number of nodes in the tree. Here we will be generating a tree 15 levels deep:

%python rdd1 = df4.rdd.map(lambda line:LabeledPoint(line[0],[line[1:]])) rdd1.take(15) model_train = DecisionTree.trainClassifier(rdd1,numClasses=2,maxDepth=15,maxBins=32,categoricalFeaturesInfo={0:2,3:8}  )   print(model_train.toDebugString()) 

Again, some of the rules look like they are redundant, so more work would be needed to transform this output to meaningful business rules.