/
evaluate_utilsPlot.py
191 lines (165 loc) · 6.5 KB
/
evaluate_utilsPlot.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
import warnings
from sklearn.exceptions import UndefinedMetricWarning
warnings.filterwarnings("ignore", category=UndefinedMetricWarning)
warnings.filterwarnings("ignore", category=UserWarning)
from sklearn.multiclass import OneVsRestClassifier
import pandas as pd
from sklearn.metrics import f1_score
from sklearn.svm import SVC
from sklearn.linear_model import LogisticRegression
from sklearn.neural_network import MLPClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import normalize as sk_normalize
from sklearn.model_selection import train_test_split
from sklearn.metrics import log_loss
import numpy as np
from collections import defaultdict
import logging
def evaluate(model, data, alg = None, classifier="lr",fast=False,ratio = None,cv=10,normalize=False,random_state = None,return_y = False):
X = model
Y = data
micros = []
macros = []
# for y,key in enumerate(data.labels.keys()):
# for index,paper in enumerate(data.labels[key]):
# if paper not in model.paper2id:
# print("paper not in model: ", paper)
# continue
# X.append(model.paper_embeddings[model.paper2id[paper]])
# Y.append(y)
print("len X: ", len(X))
print("len Y: ", len(Y))
if normalize:
X = sk_normalize(X)
scaler = StandardScaler()
X = scaler.fit_transform(X)
clf = LogisticRegression()
df = defaultdict(list)
if ratio is None:
ratio = [0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9]
for r in ratio:
if r <= 0:
continue
elif r >= 1:
break
micros = []
macros = []
for i in range(cv):
clf = LogisticRegression()
if classifier.lower() == "svm":
clf = SVC(cache_size=5000)
elif classifier.lower() == "mlp":
clf = MLPClassifier()
elif classifier.lower() == "nb":
clf = GaussianNB()
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=1-r,random_state=random_state)
clf.fit(X_train,Y_train)
prediction = clf.predict(X_test)
#lpred = clf.predict_proba(X_test)
#print("prediction shape: ", prediction[0])
#print("y_test shape: ", Y_test[0])
#print("Loss: ", log_loss(Y_test,lpred))
micro = f1_score(Y_test, prediction, average='micro')
macro = f1_score(Y_test, prediction, average='macro')
micros.append(micro)
macros.append(macro)
micros = np.mean(micros)
macros = np.mean(macros)
df["ratio"].append(r)
df["micro"].append(np.mean(micro))
df["macro"].append(np.mean(macro))
#df["alg"].append(alg)
#df["data"].append(str(data))
#df["total_samples"] = model.total_samples
#df["negative"].append(model.negative)
#df["walk_window"].append(model.walk_window)
#df["walk_probability"].append(model.walk_probability)
#df["L2"].append(model.l2)
logging.info("ratio: %.4f : f1_micro %.4f, f1_macro %.4f" % (r,micros,macros))
if fast:
if return_y:
return micros,macros,Y_test,prediction
return micros,macros
else:
return pd.DataFrame(df)
def evaluate_multilabel(model, data, alg = None, classifier="lr",fast=False,ratio = None, cv = 10, random_state = None,normalize=False):
X = []
Y = []
for pid in range(len(model.word2id)):
X.append(model.word_embeddings[pid])
Y = np.zeros((len(X),len(data.labels)))
for y,key in enumerate(data.labels.keys()):
for index,paper in enumerate(data.labels[key]):
pid = model.word2id[paper]
Y[pid][y] = 1
if normalize:
X = sk_normalize(X)
scaler = StandardScaler()
X = scaler.fit_transform(X)
df = defaultdict(list)
if ratio is None:
ratio = [0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9]
for r in ratio:
if r <= 0:
continue
elif r >= 1:
break
if classifier.lower() == 'lr':
clf = LogisticRegression()
elif classifier.lower() == "svm":
clf = SVC(cache_size=5000)
elif classifier.lower() == "mlp":
clf = MLPClassifier()
elif classifier.lower() == "nb":
clf = GaussianNB()
micros = []
macros = []
for i in range(cv):
micro,macro = evaluateNodeClassification(X,Y,1-r,clf=clf,random_state = random_state)
micros.append(micro)
macros.append(macro)
micros = np.mean(micros)
macros = np.mean(macros)
df["ratio"].append(r)
df["micro"].append(micros)
df["macro"].append(macros)
#df["alg"].append(alg)
#df["data"].append(str(data))
#df["total_samples"].append(model.total_samples)
#df["negative"].append(model.negative)
#df["walk_window"].append(model.walk_window)
#df["walk_probability"].append(model.walk_probability)
#df["L2"].append(model.l2)
logging.info("ratio: %.4f : f1_micro %.4f, f1_macro %.4f" % (r,micros,macros))
if fast:
return micros,macros
else:
return df
class TopKRanker(OneVsRestClassifier):
def predict(self, X, top_k_list):
assert X.shape[0] == len(top_k_list)
probs = np.asarray(super(TopKRanker, self).predict_proba(X))
prediction = np.zeros((X.shape[0], self.classes_.shape[0]))
for i, k in enumerate(top_k_list):
probs_ = probs[i, :]
labels = self.classes_[probs_.argsort()[-int(k):]].tolist()
for label in labels:
prediction[i, label] = 1
return prediction
def evaluateNodeClassification(X, Y, test_ratio,clf=None,random_state = None):
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=test_ratio,random_state = random_state)
try:
top_k_list = list(Y_test.toarray().sum(axis=1))
except:
top_k_list = list(Y_test.sum(axis=1))
if clf == None:
classif2 = TopKRanker(LogisticRegression())
else:
classif2 = TopKRanker(clf)
classif2.fit(X_train, Y_train)
prediction = classif2.predict(X_test, top_k_list)
print("Loss: ", log_loss(Y_test,prediction))
micro = f1_score(Y_test, prediction, average='micro')
macro = f1_score(Y_test, prediction, average='macro')
return (micro, macro)