# -*- coding:utf-8 -*-
import matplotlib
matplotlib.use('agg')
import matplotlib.pyplot as plt
import numpy as np
from matplotlib import rc
from scipy import interp
from sklearn.metrics import auc, roc_curve
from sklearn.model_selection import StratifiedKFold
from sklearn.svm import SVC
from sklearn.neighbors import KNeighborsClassifier
from sklearn.linear_model import LogisticRegression
import gcforest.data_load as load
from gcforest.gcforest import GCForest
from gcforest.utils.log_utils import get_logger
from sklearn.preprocessing import OneHotEncoder
import utils
LOGGER = get_logger('cascade_clf.lib.plot_roc_all')
def one_hot(integer_encoded):
integer_encoded = integer_encoded.reshape(len(integer_encoded), 1)
one_hot_encoder = OneHotEncoder(sparse=False)
one_hot_encoded = one_hot_encoder.fit_transform(integer_encoded)
return one_hot_encoded
save_fig = True
# activate latex text rendering
rc('text', usetex=True)
f, ax = plt.subplots(3, 1, figsize=(15, 15))
cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=0)
clf_svm = SVC(random_state=0, probability=True)
clf_knn = KNeighborsClassifier(n_neighbors=5)
clf_lg = LogisticRegression(random_state=0)
config = utils.load_json("demo_ca.json")
clf_gc = GCForest(config)
datasets = ['Cirrhosis', 'T2D', 'Obesity']
params = [(clf_lg, 'y', "LR"), (clf_svm, 'purple', 'SVM'),
(clf_knn, 'blue', 'kNN'), ('CNN', 'green', 'CNN'), (clf_gc, 'red', 'DF')]
for dataset_idx, name in enumerate(datasets):
X = None
Y = None
if dataset_idx == 0:
X, Y = load.cirrhosis_data()
elif dataset_idx == 1:
X, Y = load.t2d_data()
elif dataset_idx == 2:
X, Y = load.obesity_data()
else:
raise Exception('the dataset is not defined!!!')
gc_pred_acc = []
for idx, x in enumerate(params):
mean_fpr = np.linspace(0, 1, 100)
tprs = []
aucs = []
for train, test in cv.split(X, Y):
if idx == 3: ## CNN
Y_trans = one_hot(Y)
y_score = utils.cnn(X.iloc[train], X.iloc[test], Y_trans[train], Y_trans[test])
fpr, tpr, thresholds = roc_curve(Y[test], y_score)
v = interp(mean_fpr, fpr, tpr)
tprs.append(v)
tprs[-1][0] = 0.0
roc_auc = auc(fpr, tpr)
aucs.append(roc_auc)
elif idx == 4:
gc = x[0]
x_train, y_train, x_test, y_test = X.iloc[train], Y[train], X.iloc[test], Y[test]
x_train = x_train.values.reshape(-1, 1, len(x_train.columns))
x_test = x_test.values.reshape(-1, 1, len(x_test.columns))
X_train_enc = gc.fit_transform(x_train, y_train)
probas_ = gc.predict_proba(x_test)
fpr, tpr, thresholds = roc_curve(Y[test], probas_[:, 1])
v = interp(mean_fpr, fpr, tpr)
tprs.append(v)
tprs[-1][0] = 0.0
roc_auc = auc(fpr, tpr)
aucs.append(roc_auc)
else:
x[0].fit(X.iloc[train], Y[train])
probas_ = x[0].predict_proba(X.iloc[test])
fpr, tpr, thresholds = roc_curve(Y[test], probas_[:, 1])
v = interp(mean_fpr, fpr, tpr)
tprs.append(v)
tprs[-1][0] = 0.0
roc_auc = auc(fpr, tpr)
aucs.append(roc_auc)
mean_tpr = np.mean(tprs, axis=0)
mean_tpr[-1] = 1.0
mean_auc = auc(mean_fpr, mean_tpr)
std_auc = np.std(aucs)
if idx == 4:
label = '\\textbf{' + x[2] + '(AUC={:.3f})'.format(mean_auc) + '}'
ax[dataset_idx].plot(mean_fpr, mean_tpr, color=x[1], label=label, lw=2, alpha=.8)
else:
label = '{}' '(AUC={:.3f})'.format(x[2], mean_auc)
ax[dataset_idx].plot(mean_fpr, mean_tpr, color=x[1], label=label, lw=2, alpha=.8)
ax[dataset_idx].plot([0, 1], [0, 1], linestyle='--', lw=2, color='grey', alpha=.8)
ax[dataset_idx].set_title(name)
ax[dataset_idx].legend(loc='lower right')
ax[dataset_idx].set_ylabel('True Positive Rate')
ax[dataset_idx].set_xlabel('False Positive Rate')
ax[dataset_idx].set(adjustable='box-forced', aspect='equal')
if save_fig:
plt.savefig('output/AUC.png', bbox_inches='tight')
plt.close(f)
else:
plt.show()
Datasets
Models
