def test_RadiusNeighborsClassifier_multioutput():
"""Test k-NN classifier on multioutput data"""
rng = check_random_state(0)
n_features = 2
n_samples = 40
n_output = 3
X = rng.rand(n_samples, n_features)
y = rng.randint(0, 3, (n_samples, n_output))
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)
weights = [None, 'uniform', 'distance', _weight_func]
for algorithm, weights in product(ALGORITHMS, weights):
# Stack single output prediction
y_pred_so = []
for o in range(n_output):
rnn = neighbors.RadiusNeighborsClassifier(weights=weights,
algorithm=algorithm)
rnn.fit(X_train, y_train[:, o])
y_pred_so.append(rnn.predict(X_test))
y_pred_so = np.vstack(y_pred_so).T
assert_equal(y_pred_so.shape, y_test.shape)
# Multioutput prediction
rnn_mo = neighbors.RadiusNeighborsClassifier(weights=weights,
algorithm=algorithm)
rnn_mo.fit(X_train, y_train)
y_pred_mo = rnn_mo.predict(X_test)
assert_equal(y_pred_mo.shape, y_test.shape)
assert_array_almost_equal(y_pred_mo, y_pred_so)
def get_classifier(classifier_str):
'''
This functions maps the classifier string classifier_str to the
corresponding classifier object with the default paramers set.
'''
# SVC
if (classifier_str == 'linearsvc'):
cl = svm.LinearSVC(**svm_default_param)
elif (classifier_str == 'svc_linear'):
libsvm_default_param['kernel'] = 'linear'
cl = svm.SVC(**libsvm_default_param)
elif (classifier_str == 'svc_rbf'):
libsvm_default_param['kernel'] = 'rbf'
cl = svm.SVC(**libsvm_default_param)
# polynomial, sigmoid kernel
# nuSVC
# Nearest Neighbors (euclidian distance used by default)
elif (classifier_str == 'kn_uniform'):
kn_default_param['weights'] = 'uniform'
cl = neighbors.KNeighborsClassifier(**kn_default_param)
elif (classifier_str == 'kn_distance'):
kn_default_param['weights'] = 'distance'
cl = neighbors.KNeighborsClassifier(**kn_default_param)
elif (classifier_str == 'rn_uniform'):
rn_default_param['weights'] = 'uniform'
cl = neighbors.RadiusNeighborsClassifier(**rn_default_param)
elif (classifier_str == 'rn_distance'):
rn_default_param['weights'] = 'distance'
cl = neighbors.RadiusNeighborsClassifier(**rn_default_param)
elif (classifier_str == 'nc'):
cl = neighbors.NearestCentroid()
# LDA and QDA, priors are by default set to 1/len(class) for each class
elif (classifier_str == 'lda'):
cl = lda.LDA()
elif (classifier_str == 'qda'):
cl = qda.QDA()
# Gaussion naive bayes
# from the code it is unclear how priors are set
elif (classifier_str == 'gnb'):
cl = naive_bayes.GaussianNB()
elif (classifier_str == 'mnb'):
cl = naive_bayes.MultinomialNB()
elif (classifier_str == 'bnb'):
cl = naive_bayes.BernoulliNB()
# Decision tree
elif (classifier_str == 'dtree'):
cl = tree.DecisionTreeClassifier()
elif (classifier_str == 'rforest'):
cl = ensemble.RandomForestClassifier()
else:
# raise error if classifier not found
raise ValueError('Classifier not implemented: %s' % (classifier_str))
return (cl)
def _init_model(self, parms=None):
"""Set ML model"""
from sklearn import neighbors
if self.args.radius == 0:
if parms is not None:
model = neighbors.KNeighborsClassifier(**parms)
else:
model = neighbors.KNeighborsClassifier()
else:
if parms is not None:
model = neighbors.RadiusNeighborsClassifier(**parms)
else:
model = neighbors.RadiusNeighborsClassifier()
return 'Neighbors', model
def test_radius_neighbors_classifier_outlier_labeling():
# Test radius-based classifier when no neighbors found and outliers
# are labeled.
X = np.array([[1.0, 1.0], [2.0, 2.0], [0.99, 0.99], [0.98, 0.98],
[2.01, 2.01]])
y = np.array([1, 2, 1, 1, 2])
radius = 0.1
z1 = np.array([[1.01, 1.01], [2.01, 2.01]]) # no outliers
z2 = np.array([[1.4, 1.4], [1.01, 1.01], [2.01, 2.01]]) # one outlier
correct_labels1 = np.array([1, 2])
correct_labels2 = np.array([-1, 1, 2])
weight_func = _weight_func
for algorithm in ALGORITHMS:
for weights in ['uniform', 'distance', weight_func]:
clf = neighbors.RadiusNeighborsClassifier(radius=radius,
weights=weights,
algorithm=algorithm,
outlier_label=-1)
clf.fit(X, y)
assert_array_equal(correct_labels1, clf.predict(z1))
assert_array_equal(correct_labels2, clf.predict(z2))
def test_radius_neighbors_classifier(n_samples=40,
n_features=5,
n_test_pts=10,
radius=0.5,
random_state=0):
"""Test radius-based classification"""
rng = np.random.RandomState(random_state)
X = 2 * rng.rand(n_samples, n_features) - 1
y = ((X ** 2).sum(axis=1) < .5).astype(np.int)
y_str = y.astype(str)
weight_func = _weight_func
for algorithm in ALGORITHMS:
for weights in ['uniform', 'distance', weight_func]:
neigh = neighbors.RadiusNeighborsClassifier(radius=radius,
weights=weights,
algorithm=algorithm)
neigh.fit(X, y)
epsilon = 1e-5 * (2 * rng.rand(1, n_features) - 1)
y_pred = neigh.predict(X[:n_test_pts] + epsilon)
assert_array_equal(y_pred, y[:n_test_pts])
neigh.fit(X, y_str)
y_pred = neigh.predict(X[:n_test_pts] + epsilon)
assert_array_equal(y_pred, y_str[:n_test_pts])
def KNNski(x, r, X, Y, W='uniform'):
# we create an instance of Neighbours Classifier and fit the data.
clf = neighbors.RadiusNeighborsClassifier(r,
weights=W,
metric='seuclidean')
clf.fit(X, Y)
x = np.atleast_2d(x)
return clf.predict(x)
def _fit(cls, dataset, **hp):
classifier = neighbors.RadiusNeighborsClassifier(outlier_label=0,
radius=hp['radius'])
# Train the model using the training set
classifier.fit(dataset.train_data, dataset.train_target)
# Return score on test set
return cls._score(classifier.predict(dataset.test_data),
dataset.test_target)
def EvalKNNRadius(XT, YT, p=0.25, weights='uniform'):
xtrain, xtest, ytrain, ytest = V.train_test_split(XT,
YT,
test_size=p,
random_state=0)
knn = neighbors.RadiusNeighborsClassifier(radius=30, weights=weights)
print "Learning KNN model with " + str(weights) + " weights "
kmodel = knn.fit(xtrain, ytrain)
y_scores = kmodel.predict(xtest)
print "k-radius Train error: " + str(kmodel.score(xtrain, ytrain))
print "k-radius Test error: " + str(kmodel.score(xtest, ytest))
eval_regres(ytest, y_scores)
def radNearestNeighborsGridSeach(X, y):
param_grid = [{
'radius': np.linspace(0.1, 2, 20),
'weights': ['uniform', 'distance'],
'algorithm': ['auto', 'kd_tree'],
'outlier_label': [-1]
}]
grid_search = skgs.GridSearchCV(skn.RadiusNeighborsClassifier(),
param_grid,
cv=5)
grid_search.fit(X, y)
print 'Best Score of Grid Search: ' + str(grid_search.best_score_)
print 'Best Params of Grid Search: ' + str(grid_search.best_params_)
def rnc_tunning(Train, Test, sampling=None, scores='f1', label='FRAUDE'):
Train = pd.concat([Train, Test], axis=0, ignore_index=True)
yTrain = Train[[label]]
xTrain = Train
del xTrain[label]
if sampling == None:
pass
elif sampling == 'ALLKNN':
xTrain, yTrain = under_sampling(xTrain, yTrain)
else:
xTrain, yTrain = over_sampling(xTrain, yTrain, model=sampling)
tuned_parameters = [{'radius': list(numpy.arange(1, 101, 1)),
'weights': ['distance'], 'algorithm': ['auto'],
'leaf_size': list(numpy.arange(30, 301, 30)),
'p': [2],
'outlier_label': [-1]
}]
fileModel = GridSearchCV(neighbors.RadiusNeighborsClassifier(), param_grid=tuned_parameters, cv=10,
scoring='%s_macro' % scores)
fileModel.fit(xTrain.drop(['id_siniestro'], axis=1).values, yTrain[label].values)
print("Best parameters set found on development set:")
print()
dict_values = fileModel.best_params_
print(dict_values)
print()
print("Grid scores on development set:")
print()
means = fileModel.cv_results_['mean_test_score']
stds = fileModel.cv_results_['std_test_score']
for mean, std, params in zip(means, stds, fileModel.cv_results_['params']):
print("%0.3f (+/-%0.03f) for %r"
% (mean, std * 2, params))
radius = int(dict_values['radius'])
leaf_size = int(dict_values['leaf_size'])
df = pd.DataFrame.from_dict(dict_values, orient="index")
df.to_csv('final_files\\sup_rnc.csv', sep=';', encoding='latin1', index=False)
return radius, leaf_size, sampling
def test_radius_neighbors_classifier_zero_distance():
""" Test radius-based classifier, when distance to a sample is zero. """
X = np.array([[1.0, 1.0], [2.0, 2.0]])
y = np.array([1, 2])
radius = 0.1
z1 = np.array([[1.01, 1.01], [2.0, 2.0]])
correct_labels1 = np.array([1, 2])
weight_func = _weight_func
for algorithm in ALGORITHMS:
for weights in ['uniform', 'distance', weight_func]:
clf = neighbors.RadiusNeighborsClassifier(radius=radius,
weights=weights,
algorithm=algorithm)
clf.fit(X, y)
assert_array_equal(correct_labels1, clf.predict(z1))
def test_radius_neighbors_classifier_when_no_neighbors():
""" Test radius-based classifier when no neighbors found.
In this case it should rise an informative exception """
X = np.array([[1.0, 1.0], [2.0, 2.0]])
y = np.array([1, 2])
radius = 0.1
z1 = np.array([[1.01, 1.01], [2.01, 2.01]]) # no outliers
z2 = np.array([[1.01, 1.01], [1.4, 1.4]]) # one outlier
weight_func = _weight_func
for algorithm in ALGORITHMS:
for weights in ['uniform', 'distance', weight_func]:
clf = neighbors.RadiusNeighborsClassifier(radius=radius,
weights=weights,
algorithm=algorithm)
clf.fit(X, y)
clf.predict(z1)
assert_raises(ValueError, clf.predict, z2)
def knnRadiusTuning():
train = getTrainingData('train.csv', visualize=False)
X = train.drop(['Exited'], axis=1)
sc = StandardScaler()
X = sc.fit_transform(X)
y = train.Exited
# split training data half half
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.3,
random_state=0)
params = {
"radius": list(np.arange(5.0, 8.0, 0.1)),
"weights": ['uniform', 'distance']
}
model = neighbors.RadiusNeighborsClassifier()
grid_search_cv = GridSearchCV(model,
params,
verbose=1,
n_jobs=-1,
cv=3,
scoring='accuracy')
# print(grid_search_cv.best_params_)
grid_search_cv.fit(X_train, y_train)
print_score(grid_search_cv, X_train, y_train, X_test, y_test, train=True)
print_score(grid_search_cv, X_train, y_train, X_test, y_test, train=False)
ROC(grid_search_cv, X_train, y_train, X_test, y_test, train=True)
ROC(grid_search_cv, X_train, y_train, X_test, y_test, train=False)
results = pd.DataFrame(grid_search_cv.cv_results_)
printFullRow(results[results['rank_test_score'] == 1])
# best param setting: n_neighbors == 11/13, p ==2, weights = distance
return
# knnRadiusTuning()
# knnTuning()
def radius_neighbours(X_train, Y_train, X_test, Y_test):
try:
a=time.time()
radius=0.2 #Come back to this - need a rigorous approach
clf=neighbors.RadiusNeighborsClassifier(radius, weights='distance')
clf.fit(X_train, Y_train)
preds=clf.predict(X_test)
print
print 'Radius nearest neighbours'
print 'Time taken', time.time()-a, 's'
print 'Accuracy', sum(preds==Y_test)/(float)(len(preds))
mismatched=preds[preds!=Y_test]
print 'False Ia detection', sum(mismatched==1)/(float)(sum(preds==1))
probs=clf.predict_proba(X_test)
#fpr, tpr, auc=roc(probs, Y_test)
except ValueError:
print 'ValueError in RNN - probably due to no neighbours within radius'
fpr, tpr, auc , probs, Y_test= (None, None, None, -9999, None)
#return fpr, tpr, auc
return probs
xx, yy = np.meshgrid(np.arange(x_min, x_max, h), np.arange(y_min, y_max,
h)) #生成网格型二维数据对
Z = clf.predict(np.c_[xx.ravel(), yy.ravel()])
# Create color maps
cmap_light = ListedColormap(['#FFAAAA', '#AAFFAA', '#AAAAFF']) #给不同区域赋以颜色
cmap_bold = ListedColormap(['#FF0000', '#003300', '#0000FF']) #给不同属性的点赋以颜色
#将预测的结果在平面坐标中画出其类别区域
Z = Z.reshape(xx.shape)
plt.figure()
plt.pcolormesh(xx, yy, Z, cmap=cmap_light)
# 也画出所有的训练集数据
plt.scatter(X[:, 0], X[:, 1], c=y, cmap=cmap_bold)
plt.xlim(xx.min(), xx.max())
plt.ylim(yy.min(), yy.max())
plt.show()
clf1 = neighbors.RadiusNeighborsClassifier(10.0, weights='distance')
clf1.fit(X, y)
Z1 = clf1.predict(np.c_[xx.ravel(), yy.ravel()])
Z1 = Z1.reshape(xx.shape)
plt.figure()
plt.pcolormesh(xx, yy, Z1, cmap=cmap_light)
# 也画出所有的训练集数据
plt.scatter(X[:, 0], X[:, 1], c=y, cmap=cmap_bold)
plt.xlim(xx.min(), xx.max())
plt.ylim(yy.min(), yy.max())
plt.show()
#由图可以看出稀疏系数越小对表示结果越有利
def classify(dataSet, labels):
# 取得knn分类器
knn = neighbors.RadiusNeighborsClassifier(radius=100.0)
knn.fit(dataSet, labels)
return knn
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
''''' 创建网格以方便绘制 '''
h = .02
cmap_light = ListedColormap(['#FFAAAA', '#AAFFAA', '#AAAAFF'])
cmap_bold = ListedColormap(['#FF0000', '#00FF00', '#0000FF'])
# Plot the decision boundary. For that, we will assign a color to each
# point in the mesh [x_min, m_max]x[y_min, y_max].
x_min, x_max = X[:, 0].min() - 1, X[:, 0].max() + 1
y_min, y_max = X[:, 1].min() - 1, X[:, 1].max() + 1
xx, yy = np.meshgrid(np.arange(x_min, x_max, h), np.arange(y_min, y_max, h))
# we create an instance of Neighbours Classifier and fit the data.
''''' 训练RadiusNN分类器 '''
f = 5.0
clf = neighbors.RadiusNeighborsClassifier(radius=f,
n_neighbors=n_neighbors,
weights='distance')
clf.fit(x_train, y_train)
'''''测试结果的打印'''
answer = clf.predict(X)
# print ('X :', X)
print('answer :', answer.shape)
print('y :', y.shape)
print('avg:', np.mean(answer == y))
'''''准确率与召回率'''
# precision_recall_curve 二分类问题中使用
# precision, recall, thresholds = precision_recall_curve(y_train, clf.predict(x_train))
# answer = clf.predict_proba(X)[:,1]
print(classification_report(y, answer, target_names=['0', '1', '2']))
''''' 将整个测试空间的分类结果用不同颜色区分开'''
train_data = np.concatenate((data[0:10, ], data[50:60, ], data[100:110, ]),
axis=0)
train_label = np.concatenate(
(labels[0:10, ], labels[50:60, ], labels[100:110, ]), axis=0)
# 构造测试数据,数据垂直组合concatenate
test_data = np.concatenate((data[10:50, ], data[60:100, ], data[110:150, ]),
axis=0)
test_label_expected = np.concatenate(
(labels[10:50, ], labels[60:100, ], labels[110:150, ]), axis=0)
# 分析数据
# 创建分类器
#classifier = nb.KNeighborsClassifier(n_neighbors=3,weights='uniform',algorithm='auto')
classifier = nb.RadiusNeighborsClassifier(n_neighbors=2,
weights='uniform',
algorithm='auto')
# 训练分类器
classifier.fit(train_data, train_label)
# 预测
test_label_predicted = classifier.predict(test_data)
# 比较结果
size = len(test_label_predicted)
outer = np.zeros((size), dtype=int)
for i in range(size):
if test_label_expected[i] != test_label_predicted[i]:
outer[i] = 1
result = np.vstack((test_label_expected, test_label_predicted, outer))
def rnc_treshold(Train, Valid, Test, radius, leaf_size,
sampling=None, label='FRAUDE', beta=2):
# With beta = 2, we give the same importance to Recall and Precision
yTrain = Train[label]
xTrain = Train
del xTrain[label]
names = Train.columns.values.tolist()
fileNames = numpy.array(names)
if sampling == None:
pass
elif sampling == 'ALLKNN':
xTrain, yTrain = under_sampling(xTrain, yTrain)
else:
xTrain, yTrain = over_sampling(xTrain, yTrain, model=sampling)
min_sample_leaf = round((len(xTrain.index)) * 0.005)
fileModel = neighbors.RadiusNeighborsClassifier(radius= radius,
weights= 'distance', algorithm= 'auto',
leaf_size= leaf_size,
p= 2,
outlier_label= -1)
fileModel.fit(xTrain.values, yTrain.values)
print(np.median(fileModel.predict_proba(Valid[Valid[label] == 0].drop(label, axis=1).values)))
print(np.median(fileModel.predict_proba(Valid[Valid[label] == 1].drop(label, axis=1).values)))
tresholds = np.linspace(0.1, 1.0, 200)
scores = []
y_pred_score = fileModel.predict_proba(Valid.drop(label, axis=1).values)
y_pred_score = np.delete(y_pred_score, 0, axis=1)
print('min', y_pred_score.min())
print('max', y_pred_score.max())
for treshold in tresholds:
y_hat = (y_pred_score > treshold).astype(int)
y_hat = y_hat.tolist()
y_hat = [item for sublist in y_hat for item in sublist]
scores.append([
recall_score(y_pred=y_hat, y_true=Valid[label].values),
precision_score(y_pred=y_hat, y_true=Valid[label].values),
fbeta_score(y_pred=y_hat, y_true=Valid[label].values,
beta=2)])
scores = np.array(scores)
print('max_scores', scores[:, 2].max(), scores[:, 2].argmax())
plot.plot(tresholds, scores[:, 0], label='$Recall$')
plot.plot(tresholds, scores[:, 1], label='$Precision$')
plot.plot(tresholds, scores[:, 2], label='$F_2$')
plot.ylabel('Score')
plot.xlabel('Threshold')
plot.legend(loc='best')
plot.show()
final_tresh = tresholds[scores[:, 2].argmax()]
y_hat_test = fileModel.predict_proba(Test.drop(label, axis=1).values)
y_hat_test = np.delete(y_hat_test, 0, axis=1)
y_hat_test = (y_hat_test > final_tresh).astype(int)
y_hat_test = y_hat_test.tolist()
y_hat_test = [item for sublist in y_hat_test for item in sublist]
print('Final threshold: %.3f' % final_tresh)
print('Test Recall Score: %.3f' % recall_score(y_pred=y_hat_test, y_true=Test[label].values))
print('Test Precision Score: %.3f' % precision_score(y_pred=y_hat_test, y_true=Test[label].values))
print('Test F2 Score: %.3f' % fbeta_score(y_pred=y_hat_test, y_true=Test[label].values, beta=beta))
cnf_matrix = confusion_matrix(Test[label].values, y_hat_test)
plot_confusion_matrix(cnf_matrix, classes=['Normal', 'Anormal'], title='Confusion matrix')
featureImportance = fileModel.feature_importances_
featureImportance = featureImportance / featureImportance.max()
sorted_idx = numpy.argsort(featureImportance)
barPos = numpy.arange(sorted_idx.shape[0]) + 0.5
plot.barh(barPos, featureImportance[sorted_idx], align='center')
plot.yticks(barPos, fileNames[sorted_idx])
plot.xlabel('Variable Importance')
plot.show()
return final_tresh
# import some data to play with
iris = datasets.load_iris()
X = iris.data[:, :2] # we only take the first two features. We could
# avoid this ugly slicing by using a two-dim dataset
y = iris.target
h = .02 # step size in the mesh
# Create color maps
cmap_light = ListedColormap(['#FFAAAA', '#AAFFAA', '#AAAAFF'])
cmap_bold = ListedColormap(['#FF0000', '#00FF00', '#0000FF'])
for weights in ['uniform', 'distance']:
# we create an instance of Neighbours Classifier and fit the data.
clf = neighbors.RadiusNeighborsClassifier(n_neighbors=n_neighbors,
weights=weights)
clf.fit(X, y)
# Plot the decision boundary. For that, we will assign a color to each
# point in the mesh [x_min, m_max]x[y_min, y_max].
x_min, x_max = X[:, 0].min() - 1, X[:, 0].max() + 1
y_min, y_max = X[:, 1].min() - 1, X[:, 1].max() + 1
xx, yy = np.meshgrid(np.arange(x_min, x_max, h),
np.arange(y_min, y_max, h))
Z = clf.predict(np.c_[xx.ravel(), yy.ravel()])
# Put the result into a color plot
Z = Z.reshape(xx.shape)
plt.figure()
plt.pcolormesh(xx, yy, Z, cmap=cmap_light)