def insert_helper(self, data, current_node):
if current_node is None:
current_node = Node(data)
elif data < current_node.data:
current_node.left_child = self.insert_helper(data, current_node.left_child)
elif data > current_node.data:
current_node.right_child = self.insert_helper(data, current_node.right_child)
else:
Exception("Duplicate Entry")
return current_node
def build(self, dataset, predict_on, max_depth=5, predictors=None, min_leaf_size=5):
from Node import Node
import random
from Splitter import Splitter
node = Node()
s = Splitter()
if predictors==None:
features = list(dataset.columns)
features.remove(predict_on)
else:
features = predictors
feature = random.choice(features)
split_index, split_impurity, orig_impurity = s.split(dataset, feature)
threshold = (1.0/len(dataset))
node.split_index = split_index
if (max_depth>0) and (split_index!=0) and ((orig_impurity-split_impurity)>threshold):
node.feature = feature
node.left_child = self.build(dataset.sort_values(feature)[:split_index], predict_on, max_depth-1, predictors, min_leaf_size)
node.right_child = self.build(dataset.sort_values(feature)[split_index:], predict_on, max_depth-1, predictors, min_leaf_size)
else:
lst = list(dataset[predict_on])
node.value = sum(lst)/float(len(lst))#max(set(lst), key=lst.count)
print node.value
return node
# if __name__ == "__main__":
# import pandas as pd
# b = Builder()
# data = pd.DataFrame({'c1':[1,1,3,2],'c2':[2,3,2,2],'p':[1,1,0,0]})
# r = b.build(data, 'p')
# def print_node_message(n):
# if n.value!=None:
# print "Node " + str(n) + " has value " + str(n.value)
# else:
# print "Node is " + str(n)
# if n.left_child:
# print_node_message(n.left_child)
# else:
# print "ERROR: No left child for " + str(n)
# if n.right_child:
# print_node_message(n.right_child)
# else:
# print "ERROR: No right child for " + str(n)
# print_node_message(r)
