msaget01

Decision Tree Classification models to predict employee turnover

In this project I have attempted to create supervised learning models to assist in classifying certain employee data. The classes to predict are as follows:

• Active - the employee is still in their role
• Non-active - the employee has resigned

I pre-processed the data by removing one outlier and producing new features in Excel as the data set was small at 1056 rows. Some categorical features were also converted to numeric values in Excel. For example, Gender was originally "M" or "F", which was converted to 0 and 1 respectively. I also removed employee number as it provides no value as a feature and could compromise privacy.

After doing some research, see References, I found that the scikit-learn library does not handle categorical (string) features correctly in Decision Trees using the above approach. When added, these features provided no increase in accuracy, so I removed them. For example; Department, some departments have a highe

msaget01

msaget01

#!/usr/bin/env python
## Map based on https://cf.geekdo-images.com/images/pic38674.jpg

import networkx as nx


ttr_map = [
    ('vancouver', 'calgary', 3),
    ('vancouver', 'seattle', 1),
    ('seattle', 'calgary', 4),

msaget01

class DijkstraPath():
    def __init__(self, node_map):
        self.node_map = node_map
        self.node_length = len(node_map)
        self.used_node_list = []
        self.collected_node_dict = {}
    def __call__(self, from_node, to_node):
        self.from_node = from_node
        self.to_node = to_node
        self._init_dijkstra()

msaget01

from collections import defaultdict, deque


class Graph(object):
    def __init__(self):
        self.nodes = set()
        self.edges = defaultdict(list)
        self.distances = {}

    def add_node(self, value):