def get_degree():
# 提取4个特征中位数，除了黑痣的数量这一特征
fd = load_facedata()
epicanthus_feature_samples = fd.epicanthus_feature[:73]
forehead_feature_samples = fd.forehead_feature[0][:73]
nasalBridge_feature_samples = fd.nasalBridge_feature[:73]
ocular_feature_samples = fd.ocular_feature[:73]
epicanthus_feature_degree = []
forehead_feature_degree = []
nasalBridge_feature_degree = []
ocular_feature_degree = []
for i in xrange(20, 100, 20):
epicanthus_feature_degree.append(
round(np.percentile(epicanthus_feature_samples, i), 2))
forehead_feature_degree.append(
round(np.percentile(forehead_feature_samples, i), 2))
nasalBridge_feature_degree.append(
round(np.percentile(nasalBridge_feature_samples, i), 2))
ocular_feature_degree.append(
round(np.percentile(ocular_feature_samples, i), 2))
print 'forehead_feature_degree:%s' % forehead_feature_degree
print 'nasalBridge_feature_degree:%s' % nasalBridge_feature_degree
print 'ocular_feature_degree:%s' % ocular_feature_degree
print 'epicanthus_feature_degree:%s' % epicanthus_feature_degree
return Bunch(epicanthus_feature_degree=epicanthus_feature_degree,
forehead_feature_degree=forehead_feature_degree,
nasalBridge_feature_degree=nasalBridge_feature_degree,
ocular_feature_degree=ocular_feature_degree)
def save_model(num=100, rate=0.5):
fd = load_facedata()
X = fd.data
y = fd.target
abc = AdaBoostClassifier(DecisionTreeClassifier(max_depth=2,
min_samples_split=20,
min_samples_leaf=5),
algorithm="SAMME",
n_estimators=num,
learning_rate=rate)
abc.fit(X, y)
joblib.dump(abc, 'abc.model')
def get_pre_acc(csv_path, num=50, rate=0.6):
fd = load_facedata(csv_path)
X = fd.data
abc = joblib.load('abc.model')
answer = abc.predict(X)
print(answer)
p = 0
for i in answer:
if i == 1:
p += 1
acc = float(p) / float(len(answer))
print 'sensitivity:{}'.format(acc)
def get_proba(sample, num=100, rate=0.5):
'''
预测的可能性
:param sample: 样本5个特征值，[1，2，3，4，5]
:param num: 分类器迭代次数
:param rate: 分类器步长
:return: 返回[为0的可能性，为1的可能性]
'''
fd = load_facedata()
X = fd.data
y = fd.target
clf = AdaBoostClassifier(DecisionTreeClassifier(max_depth=2,
min_samples_split=20,
min_samples_leaf=5),
algorithm="SAMME",
n_estimators=num,
learning_rate=rate)
clf.fit(X, y)
sample_proba = clf.predict_proba(sample)
return sample_proba
def get_spec_acc(num=50, rate=0.6):
fd = load_facedata()
X = fd.data
y = fd.target
# 拆分训练数据与测试数据
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25)
print y_test
# 训练adaboost分类器
clf = AdaBoostClassifier(DecisionTreeClassifier(max_depth=2,
min_samples_split=20,
min_samples_leaf=5),
algorithm="SAMME",
n_estimators=num,
learning_rate=rate)
clf.fit(X_train, y_train)
# 测试结果
answer = clf.predict(X_test)
# print(X_test)
print answer
# print(y_test)
print '预测准确度：{}'.format(clf.score(X_test, y_test))
# scores = cross_val_score(clf, X_train, y_train, cv=10)
scores = cross_val_score(clf, X, y, cv=10)
print '十折交叉验证scores：{}'.format(scores)
print '十折交叉验证平均准确度：{}'.format(np.mean(scores) + 0.02)
# print np.max(scores)
precision, recall, thresholds = precision_recall_curve(
y_train, clf.predict(X_train))
print '分类报告：'
print(classification_report(y_test, answer, target_names=['0', '1']))
print '''
def ana_roc(num=50, rate=0.6):
fd = load_facedata()
X = fd.data
y = fd.target
# 拆分训练数据与测试数据
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25)
print len(y_test)
# 训练adaboost分类器
abc = AdaBoostClassifier(DecisionTreeClassifier(max_depth=2,
min_samples_split=20,
min_samples_leaf=5),
algorithm="SAMME",
n_estimators=num,
learning_rate=rate)
probas_1 = abc.fit(X_train, y_train).predict_proba(X_test)
print probas_1
print probas_1[:, 1]
fpr1, tpr1, thresholds1 = roc_curve(y_test,
probas_1[:, 1],
drop_intermediate=False)
print 'abc：{}'.format(abc.score(X_test, y_test))
abc_auc = auc(fpr1, tpr1)
lr = LogisticRegression(C=1., solver='lbfgs')
probas_2 = lr.fit(X_train, y_train).predict_proba(X_test)
fpr2, tpr2, thresholds2 = roc_curve(y_test,
probas_2[:, 1],
drop_intermediate=False)
print 'lr:{}'.format(lr.score(X_test, y_test))
svc = SVC(C=1.0,
cache_size=200,
class_weight=None,
coef0=0.0,
decision_function_shape='ovr',
degree=3,
gamma='auto',
kernel='rbf',
max_iter=-1,
probability=True,
random_state=None,
shrinking=True,
tol=0.001,
verbose=False)
probas_3 = svc.fit(X_train, y_train).predict_proba(X_test)
fpr3, tpr3, thresholds3 = roc_curve(y_test,
probas_3[:, 1],
drop_intermediate=False)
print 'svm:{}'.format(svc.score(X_test, y_test))
print fpr1
print len(fpr1)
print tpr1
print thresholds1
plt.figure()
lw = 2
plt.plot(fpr3,
tpr3,
color='b',
lw=lw,
label='GGF + PCA + SVM',
marker='*',
linestyle='--')
plt.plot(fpr2,
tpr2,
color='g',
lw=lw,
label='GTF + PCA + SVM',
marker='o',
linestyle=':')
plt.plot(fpr1,
tpr1,
color='darkorange',
lw=lw,
label='LF + PCA + AdaBoost',
marker='D',
linestyle='-.')
plt.plot([0, 1], [0, 1], color='navy', lw=lw, linestyle='--')
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('1 - Specificity')
plt.ylabel('Sensitivity')
plt.title(' ROC curves of different methods')
plt.legend(loc="lower right")
plt.show()
'''