def test_figure_k_acurracy(self):
""""""
#加载数据
dataset = load_files('./test_file2')
#对数据进行分词处理
datasets = []
for i in dataset.data:
datasets.append(' '.join([j for j in jieba.cut(i)])) #训练数据
cv = ShuffleSplit(n_splits=3, test_size=0.3, random_state=0)
ks = [
1, 2, 3, 5, 10, 20, 30, 50, 100, 200, 300, 500, 1000, 2000, 3000,
5000, 10000, 20000, 'all'
]
accuracys = []
for k in ks:
classifier = bayesClassifier(MultinomialNB, k=k)
clf = make_pipeline(classifier)
accuracys.append(\
average(cross_val_score(clf, datasets, dataset.target, cv=cv)))
fig, ax = plt.subplots()
ax.scatter(range(len(ks)), accuracys)
ax.set_xlabel('k')
ax.set_ylabel('accuracy')
plt.show()
print 'test_figure_k_acurracy done!'
print '-' * 70
def processData(data):
trainNum=int(len(data)*0.9)
datasets=[]
features=[]
for row in data:
datasets.append(row[1:])
features.append(row[0])
return datasets[:trainNum],features[:trainNum],datasets[trainNum:],features[trainNum:]
def processData(data):
trainNum = int(len(data) * 0.9)
datasets = []
features = []
for row in data:
datasets.append(row[1:])
features.append(row[0])
return datasets[:trainNum], features[:trainNum], datasets[
trainNum:], features[trainNum:]
def test_pipline_cross_val_score(self):
""""""
#加载数据
dataset = load_files('./test_file2')
#对数据进行分词处理
datasets = []
for i in dataset.data:
datasets.append(' '.join([j for j in jieba.cut(i)])) #训练数据
classifier = bayesClassifier(MultinomialNB, k=1000)
cv = ShuffleSplit(n_splits=3, test_size=0.3, random_state=0)
clf = make_pipeline(classifier)
print cross_val_score(clf, datasets, dataset.target, cv=cv)
print 'test_pipline_cross_val_score done!'
print '-' * 70
def make_train_test_split(X, Y, im_files, explain_files, class_names, explain_interp):
# split_data = [X_train, X_test, Y_trains, ...]
split_data = train_test_split(X, Y, im_files, explain_files, explain_interp, test_size=0.2, random_state=0)
datasets = []
for dd in range(2):
dataset = {}
dataset['X'] = split_data[dd+0]
dataset['Y'] = split_data[dd+2]
dataset['im_files'] = split_data[dd+4]
dataset['explain_files'] = split_data[dd+6]
dataset['explain_interp'] = split_data[dd+8]
dataset['class_names'] = class_names
datasets.append(dataset)
return datasets[0], datasets[1]
def load_adult(*, valid_split=0.25, path="adult.npz", device=None):
data = np.load(path)
datasets = []
for suffix in ("train", "test"):
# the 13th column is always zero
X = torch.FloatTensor(np.delete(data[f"x_{suffix}"], 13, 1))
y = torch.FloatTensor(data[f"y_{suffix}"][:, 0])
is_protected = torch.ByteTensor(
np.cast[int](data[f"attr_{suffix}"][:, 0] > 0))
X = X.to(device)
y = y.to(device)
is_protected = is_protected.to(device)
datasets.append(TensorDataset(X, y, is_protected))
train, test = datasets
in_valid = torch.rand(len(train)) < valid_split
train, valid = TensorDataset(*train[~in_valid]), TensorDataset(
*train[in_valid])
return train, valid, test
ap.add_argument("-ds", "--dataset", required=True, help="copy.txt")
ap.add_argument("-ts", "--testset", required=True, help="textcopy.txt")
ap.add_argument("-at",
"--algotype",
required=True,
help="dtc/dtr/svm/gnb/erfc/bagc/model")
ap.add_argument("-m",
"--model",
required=False,
help="dtc/dtr/svm/gnb/erfc/bagc")
args = vars(ap.parse_args())
lines_train = open(args["dataset"], 'r').readlines()
datasets = []
for n in lines_train:
datasets.append(list(map(int, n.split(' '))))
lines_test = open(args["testset"], 'r').readlines()
testset = []
for m in lines_test:
testset.append(list(map(int, m.split(' '))))
X = datasets[:-1]
target = datasets[-1]
test = testset[:-1]
test_target = testset[-1]
# X = np.c_[(0, 0, 0, 5, 5, 0, 0, 0, 2, 7, 5, 0, 1, 3, 4, 1, 0, 0, 0, 0),
# (0, 0, 0, 5, 3, 0, 0, 0, 1, 13, 3, 0, 0, 2, 5, 1, 0, 0, 0, 0),
# (0, 0, 0, 4, 5, 1, 0, 0, 3, 8, 3, 0, 0, 3, 3, 2, 0, 0, 1, 0),
# (0, 1, 1, 0, 4, 0, 0, 0, 2, 11, 5, 0, 0, 4, 3, 1, 0, 1, 0, 0),
def test_correlation_examples(N=500):
'''
Python code of correlation examples from beaucronin : https://gist.github.com/beaucronin/2509755
Python translation of examples http://en.wikipedia.org/wiki/File:Correlation_examples2.svg
Title: An example of the correlation of x and y for various distributions of (x,y) pairs
Author: Denis Boigelot
Parameters
----------
N : the number of samples
Returns
-------
'''
from numpy.random import (
uniform as runif,
multivariate_normal as rmvn,
normal as rnorm
)
from numpy import inner, linspace, array, vstack
from math import pi, cos, sin, pow, sqrt
import matplotlib.pyplot as plt
datasets = []
MI_scores = []
for corr in [1., .8, .4, 0., -.4, -.8, -1.]:
x = rmvn([0., 0.], [[1., corr], [corr, 1.]], N)
mi = MI_RenyiCC_Multi(x, type="c")
mi = compute_normMI(x, type="c")
print("MI score of two continuous linear correlated (correlation degree of %f) gaussian variables : %f" % (corr, mi))
MI_scores.append(mi)
datasets.append(x)
for phi in [0., pi/12., pi/6., pi/4., pi/2. - pi/6., pi/2. - pi/12, pi/2]:
x = rmvn([0., 0.], [[1., 1.], [1., 1.]], N)
x = rotate(phi, x)
mi = MI_RenyiCC_Multi(x, type="c")
mi = compute_normMI(x, type="c")
print("MI score of two continuous linear correlated (correlation slope of %f) gaussian variables : %f" % (phi, mi))
MI_scores.append(mi)
datasets.append(x)
a = linspace(-1, 1, N)
x = array([(x0, 4. * pow(x0 * x0 - .5, 2.) + runif(-1./3., 1./3., 1))
for x0 in a])
mi = MI_RenyiCC_Multi(x, type="c")
mi = compute_normMI(x, type="c")
print("MI score of two continuous non-linear correlated (y=4*(x0^2-0.5)^2+c) variables : %f" % mi)
MI_scores.append(mi)
datasets.append(x)
x = rotate(-pi/8., array([(x0, runif(-1., 1.)) for x0 in a]))
mi = MI_RenyiCC_Multi(x, type="c")
mi = compute_normMI(x, type="c")
print("MI score of two continuous non-linear correlated variables : %f" % mi)
MI_scores.append(mi)
datasets.append(x)
x = rotate(-pi/8, x)
mi = MI_RenyiCC_Multi(x, type="c")
mi = compute_normMI(x, type="c")
print("MI score of two continuous non-linear correlated variables : %f" % mi)
MI_scores.append(mi)
datasets.append(x)
x = array([(x0, x0 * x0 + runif(-.5, .5)) for x0 in a])
mi = MI_RenyiCC_Multi(x, type="c")
mi = compute_normMI(x, type="c")
print("MI score of two continuous non-linear correlated variables : %f" % mi)
MI_scores.append(mi)
datasets.append(x)
signs = [1. if runif() < .5 else -1. for _ in range(N)]
x = array([(x0, (x0 * x0 + runif(0., .5)) * sign)
for x0, sign in zip(a, signs)])
mi = MI_RenyiCC_Multi(x, type="c")
mi = compute_normMI(x, type="c")
print("MI score of two continuous non-linear correlated variables : %f" % mi)
MI_scores.append(mi)
datasets.append(x)
x = array([(sin(x0 * pi) + rnorm(0., .125), cos(x0 * pi) + rnorm(0., .125))
for x0 in a])
mi = MI_RenyiCC_Multi(x, type="c")
mi = compute_normMI(x, type="c")
print("MI score of two continuous non-linear correlated variables : %f" % mi)
MI_scores.append(mi)
datasets.append(x)
x = vstack((
rmvn([3., 3], [[1., 0.], [0., 1.]], round(N/4)),
rmvn([-3., 3], [[1., 0.], [0., 1.]], round(N/4)),
rmvn([-3., -3], [[1., 0.], [0., 1.]], round(N/4)),
rmvn([3., -3], [[1., 0.], [0., 1.]], round(N/4))
))
mi = MI_RenyiCC_Multi(x, type="c")
mi = compute_normMI(x, type="c")
print("MI score of two continuous non-linear correlated variables : %f" % mi)
MI_scores.append(mi)
datasets.append(x)
"""
Plot the datasets, mimicking the original plot from Wikipedia.
"""
plt.figure()
print("MI_scores:",MI_scores)
for i in range(len(datasets)):
plt.subplot(3, 7, i+1)
x = [a[0] for a in datasets[i]]
y = [a[1] for a in datasets[i]]
plt.plot(x, y, '.', markersize=1.)
plt.title("%.4f" %MI_scores[i])
# plt.axis('scaled')
plt.xticks([])
plt.yticks([])
ax = plt.gca()
ax.set_axis_off()
if i == 14:
plt.xlim([-1, 1])
plt.ylim([-1./3., 1.+1./3.])
elif i == 15:
z = sqrt(2. + sqrt(2.)) / sqrt(2.)
plt.xlim([-z, z])
plt.ylim([-z, z])
elif i == 16:
plt.xlim([-sqrt(2.), sqrt(2.)])
plt.ylim([-sqrt(2.), sqrt(2.)])
elif i == 17:
plt.xlim([-1, 1])
plt.ylim([-.5, 1.5])
elif i == 18:
plt.xlim([-1.5, 1.5])
plt.ylim([-1.5, 1.5])
elif i == 19:
plt.xlim([-1.5, 1.5])
plt.ylim([-1.5, 1.5])
elif i == 20:
plt.xlim([-7, 7])
plt.ylim([-7, 7])
else:
plt.xlim([-4, 4])
plt.ylim([-4, 4])
ax.set_aspect('equal', adjustable='datalim')
from data.load import df1pos
#dfs, dfp, dfmlle as isomap, spectral1, pca, mlle
from numpy import random
from numpy.random import randint as rndint
datalist = [dfi, dfs, dfp, dfmlle]
datasets = []
for j in datalist:
names = j.values[:, 0]
k = (j.values[:, 1:]).astype(float)
rowz = (k.shape)[0]
y = (noisy_circles[1])[0:rowz]
z = (k, y)
datasets.append(z)
#---------------------------------------------------------
colors = np.array([x for x in 'bgrcmykbgrcmykbgrcmykbgrcmyk'])
colors = np.hstack([colors] * 20)
clustering_names = [
'MiniBatchKMeans', 'AffinityPropagation', 'MeanShift',
'SpectralClustering', 'Ward', 'AgglomerativeClustering', 'DBSCAN', 'Birch'
]
plt.figure(figsize=(len(clustering_names) * 2 + 3, 9.5))
plt.subplots_adjust(left=.02,
right=.98,
bottom=.001,
scores['value'].append(metrics[i])
scores['dataset'].append(dataset)
scores_['sensitivity'].append(sens)
scores_['specificity'].append(spec)
scores_['accuracy'].append(acc)
scores_['dataset'].append(dataset)
scores_['algorithm'].append('MLP')
return scores, scores_
if __name__ == "__main__":
datasets = [('../data/features_1.npy', '../data/target_1.npy'), ('../data/mean_imputation_features_2.npy', '../data/target_1.npy' )]
datasets.append(('../data/regression_imputation_features_2.npy', '../data/target_1.npy' ))
resultnames = [ 'dataset1', 'dataset2', 'dataset3' ]
for i, (dataset, resultname) in enumerate(zip(datasets, resultnames)):
features, targets = dataset
X = np.load(features)
Y = np.load(targets)
n_classes = len(np.unique(Y))
print(n_classes)
### Stratify DATA ###
skf = StratifiedKFold(n_splits=4)