使用决策树分类器进行自动决策

Question

I'm trying to code a decision making process which otherwise require extensive IF ELSE scripting, and I'm wondering if it's possible to express the whole process in a form of root, decision and leave nodes in scikit learn's decision tree classifier.

Note that there is no training involved, just a direct user-definition of the nodes.

Answer 1

I get the intent, but a decision tree is a very particular algorithm with goals much different than yours. Depending on the kind of conditions and how many of them you have, you should first evaluate if a map (dictionary) is more appropiate. If you still think you need to go ahead with a tree structure, you should implement your own tree as @akshay-sehgal said.

You can use this simple Node class that stores the condition as a lambda, in the attribute left_condition . If left_condition is None, the node is a leaf and returns value . Otherwise it's a decision node and goes either to the left or right node, as you can see in the get_value function:

class Node:
    def __init__(left_condition = None, value = None, left = None, right= None):
    self.left_condition = left_condition
    self.value = value
    self.left = left
    self.right = right


def get_value(tree,data):
    node = tree
    while node.left_condition is not None:
        if left_condition(data):
            node = node.left
        else:
            node = node.right
    return node.value