-
Notifications
You must be signed in to change notification settings - Fork 0
/
NBAdaBoost.py
206 lines (158 loc) · 7.02 KB
/
NBAdaBoost.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
__author__ = 'manshu'
import sys
import random
import math
import NaiveBayes as nb
import csv
def findMiError(predicted_class, actual_class, tuple_weights):
error_mi = 0.0
for i in range(0, len(predicted_class)):
if predicted_class[i] != actual_class[i]:
error_mi += tuple_weights[i] * 1
else:
error_mi += tuple_weights[i] * 0
return error_mi
def assignNewTupleWeights(old_tuple_weights, eMi, predicted_class, actual_class):
new_tuple_weights = []
for i in range(0, len(predicted_class)):
new_tup_i_weight = old_tuple_weights[i] * 1.0
if predicted_class[i] == actual_class[i]:
new_tup_i_weight = old_tuple_weights[i] * (eMi / (1.0 - eMi))
new_tuple_weights.append(new_tup_i_weight)
# Normalize new_tuples_weights
old_tup_sum = sum(old_tuple_weights)
new_tup_sum = sum(new_tuple_weights)
if new_tup_sum == 0.0:
return None
normalized_tuple_weights = [tup_weight * (old_tup_sum / new_tup_sum) for tup_weight in new_tuple_weights]
return normalized_tuple_weights
def makeNewTupleIds(tuple_weights, num_training_data):
percentage_tuple_weights = [int(tup_weight * num_training_data) for tup_weight in tuple_weights]
new_tuple_ids = []
for i in range(0, num_training_data):
new_tuple_ids += [i] * percentage_tuple_weights[i]
return new_tuple_ids
def prefixScan(tuple_weights):
new_tuple_weights = []
prev_sum = 0.0
for weight in tuple_weights:
prev_sum += weight
new_tuple_weights.append(prev_sum)
return new_tuple_weights
def rangeSearch(weight_list, rand):
lo = 0
hi = len(weight_list)
while hi >= lo:
mid = (lo + hi) / 2
if rand < weight_list[mid]:
hi = mid - 1
elif rand > weight_list[mid]:
lo = mid + 1
else:
return mid
return lo
def drawRandomPD(prefixed_tuple_weights):
random_prob = random.random()
return rangeSearch(prefixed_tuple_weights, random_prob)
def formEnsembleClassifiers(training_class, training_data, max_attribute_values, k, max_run=5):
num_training_data = len(training_class)
ensemble_classifiers = []
errors_Mi = []
tuple_weights = [(1.0 / len(training_data)) for i in range(0, len(training_data))]
#tuple_ids = [i for i in range(0, len(training_class))]
for rk in range(0, k):
run_emis = []
run_classifiers = []
run_predictions = []
run_training_class = []
current_run = 0
while True:
new_training_data = []
new_training_class = []
prefixed_weights = prefixScan(tuple_weights)
for i in range(0, num_training_data):
pick_id = drawRandomPD(prefixed_weights)#random.choice(tuple_ids)
new_training_data.append(training_data[pick_id])
new_training_class.append(training_class[pick_id])
Mi = nb.makeClassifier(new_training_class, new_training_data, max_attribute_values)
predicted_class = nb.predictClass(new_training_data, Mi)
eMi = findMiError(predicted_class, new_training_class, tuple_weights)
if eMi < 0.5:# and (errors_Mi != [] and eMi < min(errors_Mi)):
ensemble_classifiers.append(Mi)
errors_Mi.append(eMi)
break
run_emis.append(eMi)
run_classifiers.append(Mi)
run_predictions.append(predicted_class)
run_training_class.append(new_training_class)
current_run += 1
#
# if current_run == max_run:
# eMi = max(run_emis)
# min_run_id = run_emis.index(eMi)
# ensemble_classifiers.append(run_classifiers[min_run_id])
# predicted_class = run_predictions[min_run_id]
# new_training_class = run_training_class[min_run_id]
# errors_Mi.append(eMi)
# break
print eMi
new_tuple_weights = assignNewTupleWeights(tuple_weights, eMi, predicted_class, new_training_class)
if new_tuple_weights == None:
rk -= 1
continue
tuple_weights = new_tuple_weights
#tuple_ids = makeNewTupleIds(tuple_weights, num_training_data)
return ensemble_classifiers, errors_Mi
def ensembleClassify(test_data, test_class, kClassifiers, kClassifiers_errors):
boosted_predicted_class = []
k_predictions = []
for i in range(0, k):
k_predictions.append(nb.predictClass(test_data, kClassifiers[i]))
for i in range(0, len(test_data)):
vote_probability = {}
for ki in range(0, k):
if k_predictions[ki][i] not in vote_probability:
vote_probability[k_predictions[ki][i]] = 0
if kClassifiers_errors[ki] == 0.0:
vote_probability[k_predictions[ki][i]] += 0
else:
vote_probability[k_predictions[ki][i]] += math.log((1.0 - kClassifiers_errors[ki]) / kClassifiers_errors[ki])
max_vote = 0
max_class = None
for class_label in vote_probability:
if vote_probability[class_label] > max_vote:
max_vote = vote_probability[class_label]
max_class = class_label
boosted_predicted_class.append(max_class)
if test_class != []:
nb.generateMeasures(test_class, boosted_predicted_class)
return boosted_predicted_class
def solveAssignment(training_file, test_file, k):
# read data from both files as it is
features, training_class, training_data = nb.readFile(training_file)
f2, ref_ids, test_data = nb.readFile(test_file)
# find max number of attributes in both the files
max_attribute, max_attribute_values = nb.findMaxNumAttributes(training_data, test_data)
# format training and test data which can be used by classifier
# training_class, training_data = nb.formatData(training_data, max_attribute)
# test_class, test_data = nb.formatData(test_data, max_attribute)
# make k classifiers from training_data and class labels using ensemble method adaboost
kClassifiers, kClassifiers_errors = formEnsembleClassifiers(training_class, training_data, max_attribute_values, k)
# print kClassifiers_errors
# predict using all the classifier built using adaboost on test data
boosted_predicted_class = ensembleClassify(training_data, training_class, kClassifiers, kClassifiers_errors)
boosted_predicted_test_class = ensembleClassify(test_data, [], kClassifiers, kClassifiers_errors)
output = []
for i in range(0, len(boosted_predicted_test_class)):
output.append([ref_ids[i], boosted_predicted_test_class[i]])
with open("output.csv", "wb") as f:
writer = csv.writer(f)
writer.writerows(output)
if __name__ == "__main__":
# if len(sys.argv) < 2:
# print "Please run the file with a training file and a test file"
# sys.exit(1)
training_file = "x.csv"#sys.argv[1]
test_file = "y.csv"#sys.argv[2]
k = 5
solveAssignment(training_file, test_file, k)