regressive_trees/tree_parallel.py at master · monkeybutter/regressive_trees · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
__author__ = 'SmartWombat'

import numpy as np
import json
from splitter_factory import SplitterFactory
from criteria_factory import CriteriaFactory
from util import angle_to_time, angle_to_date, circular_mean2, circular_mean
from datetime import date, time
from multiprocessing import Process, Manager

class Node(object):
    def __init__(self):

        self.split_var = None
        self.var_type = None
        self.score = None
        self.members = None
        self.split_values = None
        self.left_child = None
        self.right_child = None

    def get_name(self):
        return 'Node'

class Leaf(object):
    def __init__(self):

        self.value = None
        self.members = None

    def get_name(self):
        return 'Leaf'

class Tree(object):

    def tree_grower(self, data, min_leaf):

        criteria = CriteriaFactory(data.class_type, data.class_var)

        data.df = data.df[np.isfinite(data.df[data.class_var])]

        node = Node()

        best_var = None
        best_score = 0.0
        best_left = None
        best_right = None

        processes = []
        manager = Manager()
        queue = manager.Queue()

        for variable, dic in data.var_limits.iteritems():
            splitter = SplitterFactory(dic['type'], criteria)
            p = Process(target=splitter.get_split_values_queue, args=(queue, data, variable, dic['type']))
            processes.append(p)
            p.start()

        # Wait for all processes to finish
        for p in processes:
            p.join()

        for _ in processes:
            variable, type_var, score, left_df, right_df = queue.get()

            if score >= best_score:
                best_var = variable
                best_type = type_var
                best_score = score
                best_left = left_df
                best_right = right_df

        #print("Best var: " + best_var + " type " + best_type)

        node.split_var = best_var
        node.var_type = best_type
        node.split_values = best_left.var_limits[best_var]['start'], best_left.var_limits[best_var]['end'], best_right.var_limits[best_var]['start'], best_right.var_limits[best_var]['end']
        node.score = best_score
        node.members = data.df.shape[0]

        if best_left.df.shape[0]>min_leaf and np.var(best_left.df[data.class_var])!=0.0 and np.var(best_left.df[best_var])!=0.0:
            node.left_child = self.tree_grower(best_left,min_leaf)
        else:
            left_leaf = Leaf()
            #left_leaf.value = np.mean(best_left.df[data.class_var])
            left_leaf.value = self._get_leaf_value(best_left)
            left_leaf.members = best_left.df.shape[0]
            node.left_child = left_leaf

        if best_right.df.shape[0]>min_leaf and np.var(best_right.df[data.class_var])!=0.0 and np.var(best_right.df[best_var])!=0.0:
            node.right_child = self.tree_grower(best_right,min_leaf)
        else:
            right_leaf = Leaf()
            #right_leaf.value = np.mean(best_right.df[data.class_var])
            right_leaf.value = self._get_leaf_value(best_right)
            right_leaf.members = best_right.df.shape[0]
            node.right_child = right_leaf

        return node

    def _get_leaf_value(self, data):
        if data.class_type == 'linear':
            return np.mean(data.df[data.class_var])
        elif data.class_type == 'circular':
            return circular_mean2(data)
        elif data.class_type == 'date':
            print circular_mean2(data)
            print angle_to_date(circular_mean2(data))
            return angle_to_date(circular_mean2(data))
        elif data.class_type == 'time':
            print circular_mean2(data)
            print angle_to_time(circular_mean2(data))
            return angle_to_time(circular_mean2(data))
        else:
            raise Exception

    def tree_runner(self, tree, track):
        print('Node location: {}'.format(track))
        print('Split variable: {}'.format(tree.split_var))
        print('Split values: {}'.format(tree.split_value))
        print('Split score: {}'.format(tree.score))
        print('Node members: {}'.format(tree.members))

        if tree.left_child != None:
            if tree.left_child.get_name() == 'Node':
                self.tree_runner(tree.left_child, track+'L')
            elif tree.left_child.get_name() == 'Leaf':
                print('Leaf location: {}'.format(track+'Lx'))
                print('Leaf members: {}'.format(tree.left_child.members))
                print('Value {}'.format(tree.left_child.value))


        if tree.right_child != None:
            print tree.right_child.get_name()
            if tree.right_child.get_name() == 'Node':
                self.tree_runner(tree.right_child, track+'R')
            elif tree.right_child.get_name() == 'Leaf':
                print('Leaf location: {}'.format(track+'Rx'))
                print('Leaf members: {}'.format(tree.right_child.members))
                print('Value {}'.format(tree.right_child.value))

def tree_to_dict(tree, track):
    tree_dict = {}
    tree_dict['name'] = track
    tree_dict['var_name'] = tree.split_var
    tree_dict['var_type'] = tree.var_type
    tree_dict['var_limits'] = tree.split_values
    if tree.var_type == 'time':
        tree_dict['var_limits'] = angle_to_time(tree.split_values[0]).strftime("%H:%M"), angle_to_time(tree.split_values[1]).strftime("%H:%M"), angle_to_time(tree.split_values[2]).strftime("%H:%M"), angle_to_time(tree.split_values[3]).strftime("%H:%M")
    elif tree.var_type == 'date':
        tree_dict['var_limits'] = angle_to_date(tree.split_values[0]).strftime("%b %d"), angle_to_date(tree.split_values[1]).strftime("%b %d"), angle_to_date(tree.split_values[2]).strftime("%b %d"), angle_to_date(tree.split_values[3]).strftime("%b %d")
    else:
        tree_dict['var_limits'] = tree.split_values
    tree_dict['children'] = []

    if tree.left_child != None:
        if tree.left_child.get_name() == 'Node':
            #tree_dict['var_name'] = tree.left_child.split_var
            tree_dict['children'].append(tree_to_dict(tree.left_child, track+'L'))
        elif tree.left_child.get_name() == 'Leaf':
            leaf = {}
            leaf['name'] = track+'L'
            leaf['members'] = tree.left_child.members
            if isinstance(tree.left_child.value, float):
                leaf['value'] = '{0:.2f}'.format(tree.left_child.value)
            if isinstance(tree.left_child.value, time):
                leaf['value'] = tree.left_child.value.strftime('%H:%M')
            if isinstance(tree.left_child.value, date):
                leaf['value'] = tree.left_child.value.strftime('%b %d')
            tree_dict['children'].append(leaf)


    if tree.right_child != None:
        if tree.right_child.get_name() == 'Node':
            #tree_dict['var_name'] = tree.right_child.split_var
            tree_dict['children'].append(tree_to_dict(tree.right_child, track+'R'))
        elif tree.right_child.get_name() == 'Leaf':
            leaf = {}
            leaf['name'] = track+'R'
            leaf['members'] = tree.right_child.members
            if isinstance(tree.right_child.value, float):
                leaf['value'] = '{0:.2f}'.format(tree.right_child.value)
            if isinstance(tree.right_child.value, time):
                leaf['value'] = tree.right_child.value.strftime('%H:%M')
            if isinstance(tree.right_child.value, date):
                leaf['value'] = tree.right_child.value.strftime('%b %d')
            tree_dict['children'].append(leaf)

    return tree_dict
    #return json.dumps(tree_dict).replace("\"{", "{").replace("}\"", "}").replace("\\", "")