-
Notifications
You must be signed in to change notification settings - Fork 1
/
TDIDT4c.py
237 lines (179 loc) · 7.49 KB
/
TDIDT4c.py
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
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
import numpy as np
import copy
import random
import string
import graphviz as gv
import pandas as pd
class Tree(object):
def __init__(self):
self.left = None
self.right = None
self.data = None
#reads data from input file and return the list of genes, the values for each and the output for each
def readData(filename):
attr_value = []
class_label = []
fp = open(filename,'r')
lines = fp.read().split("\n")
genes = lines[0].split(',')
for rows in lines[1:]:
data = rows.split(',')
attr_value.append(data[:-1])
class_label.append(data[-1])
df = pd.read_csv(filename)
attr_value = attr_value[:-1]
for row in range(0, len(attr_value)):
for i in range(0, len(attr_value[row])):
if attr_value[row][i] == "":
temp = df.iloc[:,i].copy()
attr_value[row][i] = str(temp.median())
rez = np.array([[float(attr_value[j][i]) for j in range(len(attr_value))] for i in range(len(attr_value[0]))])
return genes[:-1], rez, class_label[:-1]
def calcEntropy(matrix):
entropy = 0
for i in list(matrix):
if i!=0:
entropy += i/float(sum(matrix)) * np.log2(1/(i/float(sum(matrix))))
#print(i,float(sum(matrix)))
return entropy
def information_gain(attr_values,class_label):
values = list(set(attr_values))
labels = list(set(class_label))
matrix = np.zeros((len(values), len(labels)))
for i in range(len(attr_values)):
matrix[values.index(attr_values[i]), labels.index(class_label[i])] += 1
h_s = calcEntropy(matrix.sum(axis=0))
h_s_a = 0
for i in range(matrix.shape[0]):
h_s_a += sum(matrix[i,:])/float(len(attr_values))*calcEntropy(matrix[i,:])
#print(sum(matrix[i,:]),float(len(attrib)))
#gain
return h_s-h_s_a
def bestSplit(attr,label_value):
unsorted = []
for i in range(len(attr)):
unsorted.append((attr[i],label_value[i]))
sortedData = sorted(unsorted, key=lambda x:x[0])
sorted_val = [val[0] for val in sortedData]
sorted_label = [val[1] for val in sortedData]
split_options = []
previous_label = sorted_label[0]
for i in range(1,len(sorted_val)):
if previous_label != sorted_label[i]:
discr_att = [0]*i+[1]*(len(sorted_val)-i)
split_options.append((i, information_gain(discr_att,sorted_label)))
previous_label = sorted_label[i]
best_split = max(split_options, key=lambda x:x[1])
if best_split[0]<=0:
avg = sorted_val[best_split[0]]
else:
avg = (float(sorted_val[best_split[0]-1]) + float(sorted_val[best_split[0]]))/2
return avg, best_split[1]
def bestAttribute(rez,class_label):
"""
Calculation of attribute with max Inf_Gain
:param rez: Data (all columns)
:param class_label: all values
:return:
"""
attrib_info_gain = []
for col_index in range(len(rez)):
avg, gain = bestSplit(rez[col_index],class_label)
attrib_info_gain.append([avg,gain,col_index]) #col_index is the number of column
best_attribute = max(attrib_info_gain, key=lambda x:x[1])
return best_attribute
def splitData(best_attribute,data,class_label):
right_data = []
left_data = []
right_label = []
left_label = []
for i in range(0, len(data[best_attribute[2]])-1):
if best_attribute[0] < data[best_attribute[2]][i]:
right_data.append(data[:, i])
right_label.append(class_label[i])
else:
left_data.append(data[:, i])
left_label.append(class_label[i])
return np.array(right_data).transpose(), right_label, np.array(left_data).transpose() , left_label
def tdidt(genes,data,class_label,depth,tree):
if depth >= 3 or len(data) == 0 or class_label.count(0) == len(data) or class_label.count(1) == len(data):
tree['gene'] = 'leaf'
tree['children'] = []
tree['value'] = [class_label.count(0), class_label.count(1)]
tree['label'] = max(set(class_label), key = class_label.count)
dims = data.shape
tree['data samples'] = dims[1]
return tree
else:
best_attribute = bestAttribute(data,class_label)
tree['decision'] = best_attribute[0]
tree['gain'] = best_attribute[1]
tree['id'] = best_attribute[2]
tree['gene'] = genes[best_attribute[2]]
dims = data.shape
tree['data samples'] = dims[1]
tree['value'] = [class_label.count(0), class_label.count(1)]
tree['children'] = []
right_data,right_label,left_data,left_label = splitData(best_attribute,data,class_label)
#returns one column!!
tree['children'].append({})
tdidt(genes,left_data,left_label,depth+1,tree['children'][-1])
tree['children'].append({})
tdidt(genes,right_data,right_label,depth+1,tree['children'][-1])
def tree_dot(outfile, tree):
tree_dot = gv.Digraph(format='svg',engine='dot')
traversal(tree, 'root', tree_dot)
f = open(outfile,'w+')
f.write(tree_dot.source)
f.close()
def traversal(current, position, tree_dot):
if current['gene'] == 'leaf':
tree_dot.attr('node', shape='box')
name = position + str(random.choice(string.ascii_lowercase + string.digits))
tree_dot.node(name, ''' samples = %(samples)d \n healthy = %(healthy)d, trisomic = %(trisomic)d \n class = %(class)d''' % {'samples': current['data samples'], 'healthy':current['value'][0], 'trisomic': current['value'][1], 'class':current['label']})
tree_dot.edge(position, name)
else:
tree_dot.attr('node', shape='box')
name = current['gene'] + '_' + str(current['data samples'])
tree_dot.node(name = name, label = '''%(property_name)s >= %(dec)f \n samples = %(samples)d \n healthy = %(healthy)d, trisomic = %(trisomic)d''' % {'property_name': current['gene'], 'dec': current['decision'], 'samples':current['data samples'],'healthy':current['value'][0], 'trisomic': current['value'][1]})
if position != 'root':
tree_dot.edge(position, name)
for children in current['children']:
traversal(children, name, tree_dot)
def predict(tree, data):
predicted_labels = []
for row in data:
predicted_labels.append(classify(tree, row))
return predicted_labels
def classify(tree, data_row):
tree_copy = tree
while True:
if tree_copy['gene'] == 'leaf':
return tree_copy['label']
elif data_row[tree_copy['id']] > tree_copy['decision']:
tree_copy = tree_copy['children'][1]
continue
else:
tree_copy = tree_copy['children'][0]
def accuracy(pred, true):
correct = 0
for pr, tr in zip(pred, true):
if pr == float(tr):
correct = correct + 1
return float(correct)/len(pred)*100
print('Reading Data')
genes,all_data,class_label = readData("gene_expression_with_missing_values.csv")
print('Building tree')
tree = {}
tdidt(genes,all_data,[float(i) for i in class_label],0,tree)
tree_dot('tree4c.dot', tree)
print(tree)
print('Loading Testing Data')
file = [line.strip().split(',') for line in open('gene_expression_test.csv','r')]
genes = file[0][:-1]
test_data = [[float(el) for el in row[:-1]] for row in file[1:]]
test_label = [float(row[-1]) for row in file[1:]]
print('Testing Tree')
predicted = predict(tree, test_data)
accuracy = accuracy(predicted, test_label)
print('Decision Tree was ' + str(accuracy) + ' accurate')