def create_random_decision_tree(context, depth):
if 0 == depth: return LeafNode(choice(context.outcomes))
else:
rule, num_outcomes = new_rule(context.attributes)
children = [
create_random_decision_tree(context, depth-1) for i in xrange(num_outcomes)
]
return DecisionNode(rule, children)
def replace_rule(decision_node, context, depth = 0):
"Replace the rule in the node with a completely new random rule"
decision_node.rule, num_outcomes = new_rule(context.attributes)
while num_outcomes > len(decision_node.children): # add children if needed
decision_node.children.append(create_random_decision_tree(context, depth))
while num_outcomes < len(decision_node.children): # remove children if too many
decision_node.children.pop()
shuffle(decision_node.children) # randomise order
def insert_decision(decision_node, context, depth = 0):
"Insert an extra decision in the tree"
child_to_replace = choice(decision_node.children)
idx_to_replace = decision_node.children.index(child_to_replace)
rule, num_outcomes = new_rule(context.attributes)
children = [ child_to_replace ]
while num_outcomes > len(children): # add other children
children.append(create_random_decision_tree(context, depth))
decision_node.children[idx_to_replace] = DecisionNode(rule, children)
