-
Notifications
You must be signed in to change notification settings - Fork 0
/
analysis_ml.py
173 lines (138 loc) · 7.56 KB
/
analysis_ml.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
import os
import re
import pickle
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.metrics import roc_auc_score, roc_curve, auc
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import cross_validate
from sklearn.model_selection import train_test_split
from sklearn.metrics import DetCurveDisplay, RocCurveDisplay
from sklearn.model_selection import StratifiedKFold
from sklearn.metrics import confusion_matrix
def import_excel_files(path, regex):
files_list = [file for file in os.listdir(path) if
os.path.isfile(os.path.join(path, file))]
r = re.compile(regex) # by regex ^train_*
files_list = list(filter(r.match, files_list))
# concatenate all models results
results = pd.DataFrame()
for file in files_list:
file = path + file
df = pd.read_excel(file, index_col=0, header=None)
df.set_axis(df.loc['TARGET'], axis='columns', inplace=True)
df = df.sort_values(by='roc_auc', axis=1, ascending=False) # sort for drop cols except max
df = df.loc[:, ~df.columns.duplicated()]
results = pd.concat([results, df], axis=1)
results = results.T.reset_index(drop=True).T # drop column names
results.to_excel('results\\all_models_results.xlsx') # export data
return results
if __name__ == '__main__':
# style for plots
plt.style.use('seaborn')
sns.set_context("talk") # \ "paper" \ "poster" \ "notebook"
sns.set_style("whitegrid")
cmap = plt.cm.get_cmap('cool')
# import
path = 'results\\' # files path
regex = '^train_*' # files' names
results = import_excel_files(path, regex)
all_scores = pd.DataFrame(index=['TARGET', 'Model']) # empty score file create
for target in results.loc['TARGET'].unique(): # targets loop
fig, ax = plt.subplots(figsize=(12, 6)) # ROC curve for each target
colors_num = results.loc['Model'].unique().shape[0] # color for each line from cmap
for n, model in enumerate(results.loc['Model'].unique()): # models loop
metric = results.loc['roc_auc', (results.loc['Model'] == model) &
(results.loc['TARGET'] == target)].values
if metric.size == 0: # if no data
print('\033[91m Warning: No data for ', target, '|', model, '\033[0m')
else: # load model and datasets
file_name = target.replace(" ", "_") + '_' + \
model + '_roc_auc_' + str(metric[0])
file = 'models\\' + file_name + '_model.sav'
clf = pickle.load(open(file, 'rb'))
file = 'models\\' + file_name + '_target.sav'
y = pd.read_pickle(file)
file = 'models\\' + file_name + '_predicts.sav'
X = pd.read_pickle(file)
scores = cross_validate(clf, X, y, cv=5, scoring=[ # model evaluation on CV5
'roc_auc',
'accuracy',
'balanced_accuracy',
'average_precision',
'precision',
'recall',
'f1',
'f1_micro',
'f1_macro',
'neg_log_loss'])
scores = pd.Series(scores).apply(lambda x: round(np.mean(x), 3)) # save average score
scores['TARGET'] = target
scores['Model'] = model
print(file_name, '\tTest_score\t', scores['test_roc_auc'])
cv = StratifiedKFold(n_splits=5) # CV5 for ROC plotting
X = np.array(X)
y = np.array(y)
mean_fpr = np.linspace(0, 1, 1000)
tprs = []
aucs = []
tn = [None] * 5 # * 5 - CV
fp = [None] * 5
fn = [None] * 5
tp = [None] * 5
y_pos = [None] * 5
y_neg = [None] * 5
for i, (train, test) in enumerate(cv.split(X, y)):
clf.fit(X[train], y[train])
tn[i], fp[i], fn[i], tp[i] = confusion_matrix(clf.predict(X[test]), y[test]).ravel()
y_pos[i] = np.sum(y[test] == 1)
y_neg[i] = np.count_nonzero(y[test] == 0)
viz = RocCurveDisplay.from_estimator(clf, X[test], y[test],
name='', alpha=0, lw=1,
ax=ax, label='_nolabel_')
interp_tpr = np.interp(mean_fpr, viz.fpr, viz.tpr)
interp_tpr[0] = 0.0
tprs.append(interp_tpr)
aucs.append(viz.roc_auc)
mean_tpr = np.mean(tprs, axis=0)
mean_tpr[-1] = 1.0
mean_auc = auc(mean_fpr, mean_tpr)
print('roc_auc from ROC curve', mean_auc)
line = ax.plot(mean_fpr, mean_tpr, lw=2,
label ='{string} Mean AUC = {auc:.2f} $\pm$ {std:.2f}'.
format(string=model, auc=mean_auc, std=np.std(aucs)))
line[-1].set_color(cmap(n/colors_num)) # replace line's color
scores['True Positive'] = tp
scores['Y positive number'] = y_pos
scores['True Negative'] = tn
scores['Y negative number'] = y_neg
scores['False Positive'] = fp
scores['False Negative'] = fn
scores['roc_auc2'] = mean_auc
all_scores = pd.concat([all_scores, scores], axis=1)
#break
ax.plot([0, 1], [0, 1], linestyle="--", lw=1, color="gray", label='_nolabel_', alpha=0.8)
ax.set(xlim=[-0, 1.0], ylim=[-0, 1.0], title=target + ' | Mean ROC curve')
ax.legend(bbox_to_anchor=(1.05, 1), loc='upper left', borderaxespad=0.)
plt.tight_layout()
file = 'results\\Model_comparison_for_' + target + '.tiff'
fig.savefig(file, dpi=200)
#break
all_scores.to_excel('results\\all_scores.xlsx', header=False)
# TODO: export best results + other metrics
# TODO: tp fp square https://towardsdatascience.com/metrics-and-python-ii-2e49597964ff
# TODO: IMPORTANCE
# feature_scores = pd.Series(model.feature_importances_, index=predicts.columns).sort_values(ascending=False)
# feature_scores = feature_scores[feature_scores != 0]
# fig, ax = plt.subplots(figsize=(10, 5))
# ax = sns.barplot(x=feature_scores, y=feature_scores.index)
# ax.set_title("Visualize feature scores of the features")
# ax.set_yticklabels(feature_scores.index)
# ax.set_xlabel("Feature importance score")
# ax.set_ylabel("Features")
# plt.tight_layout()
# plt.show()
# TODO: Tree graph https://russianblogs.com/article/5797349374/
### "Linear SVM": make_pipeline(StandardScaler(), LinearSVC(C=0.025)),