-
Notifications
You must be signed in to change notification settings - Fork 1
/
train_classifier.py
141 lines (110 loc) · 5.96 KB
/
train_classifier.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
import pandas as pd
import re
from sklearn.model_selection import train_test_split, cross_val_predict
from sklearn.linear_model import RidgeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report, confusion_matrix
from sklearn.externals import joblib
import numpy as np
import seaborn as sb
import matplotlib.pyplot as plt
from collections import Counter
data = pd.read_csv(r'dataset\\all_data\\result.csv')
expert_labels = []
for elem in data['filename']:
expert_labels.append(int(re.findall('\d+', elem)[0]))
data = data.drop(labels=['filename'], axis=1)
two_class = False
if two_class:
for j in range(len(expert_labels)):
if expert_labels[j] == 4:
expert_labels[j] = 3
to_be_dropped = []
normalization = True
if normalization:
for feature in data:
delimiter = data[feature].max()-data[feature].min()
if delimiter == 0:
to_be_dropped.append(feature)
else:
data[feature] = (data[feature] - data[feature].mean())/(delimiter)
formal = data.loc[:, ['avg_len_in_chars', 'len_in_chars', 'len_in_words']]
data = data.drop(labels=['len_in_chars', 'len_in_words'], axis=1)
sylls_morphs_accents = data.loc[: ,['stressed_first_v_W', 'c_in_the_end_W', 'c_in_the_beginning_W', 'two_syl_open_syls_W',
'three_syl_open_syls_W', 'one_syl_W', 'two_syl_W', 'one_syl_cvc_W', 'one_syl_begin_cc_W',
'two_syl_begin_cc_W', 'two_syl_1th_stressed_W', 'three_syl_2nd_stressed_W', 'two_syl_2nd_stressed_W',
'three_syl_1th_stressed_W', 'three_syl_cv_pattern_W', 'four_syl_cv_pattern_W', 'nom_W',
'acc_W', 'dat_W', 'abl_W', 'verbs_pers_S', 'parenth_S', 'one_syl_end_cc_W', 'two_syl_middle_cc_W',
'three_syl_begin_cc_W', 'three_syl_middle_cc_W', 'three_syl_end_cc_W', 'four_syl_cc_on_the_edge_W',
'five_syl_cv_pattern_W', 'adv_W', 'gen_W', 'ins_W', 'numeral_W', 'a_pro_W', 'coord_conjs_num_S',
's_pro_S', 'three_syl_3rd_stressed_W', 'three_syl_cc_on_the_edge_W', 'five_syl_cc_on_the_edge_W',
'alt_conjs_num_S', 'abstr_nouns_rate_S', 'foreign_W']]
lexs = data.loc[:, ['rare_obsol_W', 'avg_W_freq_S', 'avg_W_Rs_S', 'avg_W_Ds_S', 'avg_W_Docs_S', 'oov_words_rate_S',
'N_top_200_rate_S', 'N_top_400_rate_S', 'N_top_600_rate_S', 'N_top_800_rate_S', 'N_top_1000_rate_S',
'V_top_200_rate_S', 'V_top_400_rate_S', 'V_top_600_rate_S', 'V_top_800_rate_S', 'V_top_1000_rate_S',
'A_top_200_rate_S', 'A_top_400_rate_S', 'A_top_600_rate_S', 'A_top_800_rate_S', 'A_top_1000_rate_S']]
syntax = data.loc[:, ['sent_simple_S', 'sent_two_homogen_S', 'sent_three_homogen_S', 'no_predic_S', 'sent_complic_soch_S',
'sent_complic_depend_S', 'inverse_S']]
if normalization:
print(to_be_dropped)
if to_be_dropped:
data = data.drop(labels=to_be_dropped, axis=1)
sylls_morphs_accents = sylls_morphs_accents.drop(labels=['foreign_W'], axis=1)
#data = data.loc[:, ['inverse_S', 'abstr_nouns_rate_S', 'parenth_S', 'acc_W', 'two_syl_W', 'three_syl_cc_on_the_edge_W',
# 'avg_len_in_chars', 'gen_W', 'sent_complic_depend_S', 'three_syl_3rd_stressed_W', 'two_syl_middle_cc_W',
# 'a_pro_W', 'five_syl_cc_on_the_edge_W', 'one_syl_W', 'dat_W', 'three_syl_2nd_stressed_W', 'N_top_800_rate_S']]
for data_type in [data]:#[formal, sylls_morphs_accents, lexs, syntax, data]:
X_train, X_test, y_train, y_test = train_test_split(data_type, expert_labels, test_size=1/3, random_state=42,
stratify=expert_labels)
"""print('X_train:', Counter(expert_labels))
print('y_train:', Counter(y_train))
print('y_test:', Counter(y_test))"""
#classifier = RidgeClassifier(random_state=42)#,alpha=0.8, fit_intercept=True, normalize=True)
classifier = RandomForestClassifier(random_state=42, n_jobs=-1, max_features=0.72,
class_weight='balanced', n_estimators=48)
"""predicted = cross_val_predict(classifier, data_type, expert_labels, cv = 3)
print(classification_report(expert_labels, predicted))
conf_matrix = confusion_matrix(expert_labels, predicted)
print(conf_matrix)"""
classifier.fit(X_train, y_train)
save_model = False
if save_model:
joblib.dump(classifier, 'trained_model')
y_pred = classifier.predict(X_test)
print(classification_report(y_test, y_pred))
conf_matrix = confusion_matrix(y_pred, y_test)
print(conf_matrix)
spec_score = 0
full_match = 0
part_match = 0
penalty = 0
for i in range(len(y_pred)):
if y_pred[i] == y_test[i]:
spec_score += 1
full_match += 1
elif abs(y_pred[i]-y_test[i]) == 1:
if y_test[i] == 3:
pass
else:
spec_score += 0.25
part_match += 1
elif abs(y_pred[i]-y_test[i]) == 2:
spec_score -= 0.5
penalty += 1
print('our_score', spec_score/len(y_pred), spec_score)
print(full_match, part_match, penalty)
print_heatmap = False
if print_heatmap:
plt.figure(figsize = (10,7))
sb.heatmap(conf_matrix, annot=True, xticklabels=['2 класс true','3 класс true','4 класс true'],
yticklabels=['2 класс pred','3 класс pred','4 класс pred'])
plt.show()
print_importances = False
if print_importances:
importances = classifier.feature_importances_
features_names = [name for name in data_type]
indices = np.argsort(importances)[::-1]
# print the feature ranking
print("Feature ranking:")
for f in range(data_type.shape[1]):
print("%d. %s (%f)" % (f + 1, features_names[indices[f]], importances[indices[f]]))