/
example.py
217 lines (195 loc) · 11.5 KB
/
example.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
"""
This is an example script that will apply k-cross-validation on all datasets with a load function in
`data.load_datasets` and for all implemented tree constructors, ensemble techniques and GENESIM. In the end,
a confusion matrices will be stored at path `output/dataset_name_CVk.png` and the average model complexity and
computational time required for each of the algorithms will be printed out.
Written by Gilles Vandewiele in commission of IDLab - INTEC from University Ghent.
"""
import time
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis, QuadraticDiscriminantAnalysis
from sklearn.grid_search import GridSearchCV
from sklearn.metrics import confusion_matrix
from sklearn.cross_validation import StratifiedKFold, KFold
import matplotlib.pyplot as plt
import numpy as np
from sklearn.neural_network import MLPClassifier
import constructors.ISM
from constructors.ensemble import RFClassification, XGBClassification, bootstrap
from constructors.genesim import GENESIM
from constructors.inTrees import inTreesClassifier
from constructors.treeconstructor import QUESTConstructor, GUIDEConstructor, C45Constructor, CARTConstructor
from data.load_all_datasets import load_all_datasets
from decisiontree import DecisionTree
if __name__ == "__main__":
algorithms = {QUESTConstructor().get_name(): QUESTConstructor(),
GUIDEConstructor().get_name(): GUIDEConstructor(),
CARTConstructor().get_name(): CARTConstructor(),
C45Constructor().get_name(): C45Constructor(),
RFClassification().get_name(): RFClassification(),
XGBClassification().get_name(): XGBClassification()
}
genesim = GENESIM()
inTrees_clf = inTreesClassifier()
NR_FOLDS = 5
for dataset in load_all_datasets():
df = dataset['dataframe']
label_col = dataset['label_col']
feature_cols = dataset['feature_cols']
conf_matrices, avg_nodes, times = {}, {}, {}
for algorithm in algorithms:
conf_matrices[algorithm] = []
avg_nodes[algorithm] = []
times[algorithm] = []
conf_matrices['GENESIM'], avg_nodes['GENESIM'], times['GENESIM'] = [], [], []
conf_matrices['ISM'], avg_nodes['ISM'], times['ISM'] = [], [], []
conf_matrices['inTrees'], avg_nodes['inTrees'], times['inTrees'] = [], [], []
skf = StratifiedKFold(df[label_col], n_folds=NR_FOLDS, shuffle=True, random_state=None)
for fold, (train_idx, test_idx) in enumerate(skf):
print 'Fold', fold+1, '/', NR_FOLDS, 'for dataset', dataset['name']
train = df.iloc[train_idx, :].reset_index(drop=True)
X_train = train.drop(label_col, axis=1)
y_train = train[label_col]
test = df.iloc[test_idx, :].reset_index(drop=True)
X_test = test.drop(label_col, axis=1)
y_test = test[label_col]
for algorithm in algorithms:
print algorithm
start = time.time()
clf = algorithms[algorithm].construct_classifier(train, feature_cols, label_col)
end = time.time()
times[algorithm].append(end-start)
predictions = clf.evaluate_multiple(X_test).astype(int)
conf_matrix = confusion_matrix(y_test, predictions)
print conf_matrix
diagonal_sum = sum(
[conf_matrix[i][i] for i in range(len(conf_matrix))])
norm_diagonal_sum = sum(
[float(conf_matrix[i][i]) / float(sum(conf_matrix[i])) for i in
range(len(conf_matrix))])
total_count = np.sum(conf_matrix)
print 'Accuracy:', float(diagonal_sum) / float(total_count)
print 'Balanced accuracy:', float(norm_diagonal_sum) / float(conf_matrix.shape[0])
conf_matrices[algorithm].append(confusion_matrix(y_test, predictions))
if type(clf) is DecisionTree:
avg_nodes[algorithm].append(clf.count_nodes())
else:
avg_nodes[algorithm].append(clf.nr_clf)
_constructors = [CARTConstructor(), QUESTConstructor(), GUIDEConstructor()]
print 'inTrees'
start = time.time()
orl = inTrees_clf.construct_rule_list(train, label_col, _constructors, nr_bootstraps=25)
end = time.time()
times['inTrees'].append(end-start)
predictions = orl.evaluate_multiple(X_test).astype(int)
conf_matrices['inTrees'].append(confusion_matrix(y_test, predictions))
conf_matrix = confusion_matrix(y_test, predictions)
print conf_matrix
diagonal_sum = sum(
[conf_matrix[i][i] for i in range(len(conf_matrix))])
norm_diagonal_sum = sum(
[float(conf_matrix[i][i]) / float(sum(conf_matrix[i])) for i in
range(len(conf_matrix))])
total_count = np.sum(conf_matrix)
correct = 0
for i in range(len(conf_matrix)):
correct += conf_matrix[i][i] + conf_matrix[i][max(i - 1, 0)] * ((i - 1) >= 0) + \
conf_matrix[i][min(i + 1, len(conf_matrix[i]) - 1)] * ((i + 1) <= len(conf_matrix[i]) - 1)
# print 'Accuracy [-1, +1]:', float(correct) / float(total_count)
print 'Accuracy:', float(diagonal_sum) / float(total_count)
print 'Balanced accuracy:', float(norm_diagonal_sum) / float(conf_matrix.shape[0])
avg_nodes['inTrees'].append(len(orl.rule_list))
print 'ISM'
start = time.time()
ism_tree = constructors.ISM.ism(bootstrap(train, label_col, _constructors, boosting=True, nr_classifiers=5),
train, label_col, min_nr_samples=1, calc_fracs_from_ensemble=False)
ism_pruned = ism_tree.cost_complexity_pruning(X_train, y_train, 'ism', ism_constructors=_constructors,
ism_calc_fracs=False, n_folds=3, ism_nr_classifiers=5,
ism_boosting=True)
end = time.time()
times['ISM'].append(end - start)
predictions = ism_pruned.evaluate_multiple(X_test).astype(int)
conf_matrices['ISM'].append(confusion_matrix(y_test, predictions))
avg_nodes['ISM'].append(ism_pruned.count_nodes())
conf_matrix = confusion_matrix(y_test, predictions)
print conf_matrix
diagonal_sum = sum(
[conf_matrix[i][i] for i in range(len(conf_matrix))])
norm_diagonal_sum = sum(
[float(conf_matrix[i][i]) / float(sum(conf_matrix[i])) for i in
range(len(conf_matrix))])
total_count = np.sum(conf_matrix)
correct = 0
for i in range(len(conf_matrix)):
correct += conf_matrix[i][i] + conf_matrix[i][max(i - 1, 0)] * ((i - 1) >= 0) + \
conf_matrix[i][min(i + 1, len(conf_matrix[i]) - 1)] * ((i + 1) <= len(conf_matrix[i]) - 1)
# print 'Accuracy [-1, +1]:', float(correct) / float(total_count)
print 'Accuracy:', float(diagonal_sum) / float(total_count)
print 'Balanced accuracy:', float(norm_diagonal_sum) / float(conf_matrix.shape[0])
avg_nodes['ISM'].append(ism_pruned.count_nodes())
print 'GENESIM'
# train_gen = train.rename(columns={'Class': 'cat'})
start = time.time()
genetic = genesim.genetic_algorithm(train, label_col, _constructors, seed=None, num_iterations=40,
num_crossovers=15, population_size=250, val_fraction=0.4, prune=True,
max_samples=3, tournament_size=15, nr_bootstraps=40)
end = time.time()
times['GENESIM'].append(end - start)
predictions = genetic.evaluate_multiple(X_test).astype(int)
conf_matrices['GENESIM'].append(confusion_matrix(y_test, predictions))
conf_matrix = confusion_matrix(y_test, predictions)
print conf_matrix
diagonal_sum = sum(
[conf_matrix[i][i] for i in range(len(conf_matrix))])
norm_diagonal_sum = sum(
[float(conf_matrix[i][i]) / float(sum(conf_matrix[i])) for i in
range(len(conf_matrix))])
total_count = np.sum(conf_matrix)
correct = 0
for i in range(len(conf_matrix)):
correct += conf_matrix[i][i] + conf_matrix[i][max(i - 1, 0)] * ((i - 1) >= 0) + \
conf_matrix[i][min(i + 1, len(conf_matrix[i]) - 1)] * ((i + 1) <= len(conf_matrix[i]) - 1)
# print 'Accuracy [-1, +1]:', float(correct) / float(total_count)
print 'Accuracy:', float(diagonal_sum) / float(total_count)
print 'Balanced accuracy:', float(norm_diagonal_sum) / float(conf_matrix.shape[0])
avg_nodes['GENESIM'].append(genetic.count_nodes())
print times
print avg_nodes
fig = plt.figure()
fig.suptitle('Accuracy on ' + dataset['name'] + ' dataset using ' + str(NR_FOLDS) + ' folds', fontsize=20)
counter = 0
conf_matrices_mean = {}
for key in conf_matrices:
conf_matrices_mean[key] = np.zeros(conf_matrices[key][0].shape)
for i in range(len(conf_matrices[key])):
conf_matrices_mean[key] = np.add(conf_matrices_mean[key], conf_matrices[key][i])
cm_normalized = np.around(
conf_matrices_mean[key].astype('float') / conf_matrices_mean[key].sum(axis=1)[:,
np.newaxis], 4)
diagonal_sum = sum(
[conf_matrices_mean[key][i][i] for i in range(len(conf_matrices_mean[key]))])
norm_diagonal_sum = sum(
[conf_matrices_mean[key][i][i]/sum(conf_matrices_mean[key][i]) for i in range(len(conf_matrices_mean[key]))])
total_count = np.sum(conf_matrices_mean[key])
print key
print conf_matrices_mean[key]
correct = 0
for i in range(len(conf_matrices_mean[key])):
correct += conf_matrices_mean[key][i][i] + conf_matrices_mean[key][i][max(i - 1, 0)] * ((i - 1) >= 0) + \
conf_matrices_mean[key][i][min(i + 1, len(conf_matrices_mean[key][i]) - 1)] * ((i + 1) <= len(conf_matrices_mean[key][i]) - 1)
print 'Accuracy [-1, +1]:', float(correct) / float(total_count)
print 'Accuracy:', float(diagonal_sum) / float(total_count)
print 'Balanced accuracy:', float(norm_diagonal_sum) / conf_matrices_mean[key].shape[0]
ax = fig.add_subplot(2, np.math.ceil(len(conf_matrices) / 2.0), counter + 1)
cax = ax.matshow(cm_normalized, cmap=plt.cm.Blues, vmin=0.0, vmax=1.0)
ax.set_title(key + '(' + str(sum(avg_nodes[key])/len(avg_nodes[key])) + ')', y=1.08)
for (j, i), label in np.ndenumerate(cm_normalized):
ax.text(i, j, label, ha='center', va='center')
if counter == len(conf_matrices) - 1:
fig.colorbar(cax, fraction=0.046, pad=0.04)
counter += 1
F = plt.gcf()
Size = F.get_size_inches()
F.set_size_inches(Size[0] * 2, Size[1] * 1.75, forward=True)
plt.show()
rand_nr = str(int(10000*np.random.rand()))
plt.savefig('output/' + dataset['name'] + '_CV'+str(NR_FOLDS)+'.png', bbox_inches='tight')