def test_evaluate_equal_dim_and_num_lt(self):
x1 = np.arange(3, 10, 2)
x2 = np.arange(3, 8, 2)
kde = mlab.GaussianKDE(x1)
y_expected = [0.08797252, 0.11774109, 0.11774109]
y = kde.evaluate(x2)
np.testing.assert_array_almost_equal(y, y_expected, 7)
def test_wrong_bw_method(self):
"""Test the error message that should be called when bw is invalid."""
np.random.seed(8765678)
n_basesample = 50
data = np.random.randn(n_basesample)
with pytest.raises(ValueError):
mlab.GaussianKDE(data, bw_method="invalid")
def test_kde_integer_input(self):
"""Regression test for #1181."""
x1 = np.arange(5)
kde = mlab.GaussianKDE(x1)
y_expected = [0.13480721, 0.18222869, 0.19514935, 0.18222869,
0.13480721]
np.testing.assert_array_almost_equal(kde(x1), y_expected, decimal=6)
def test_evaluate_dim_and_num(self):
"""Tests if evaluated against a one by one array"""
x1 = np.arange(3, 10, 2)
x2 = np.array([3])
kde = mlab.GaussianKDE(x1)
y_expected = [0.08797252]
y = kde.evaluate(x2)
np.testing.assert_array_almost_equal(y, y_expected, 7)
def fit_density(self, X):
if frequency_interpolation_type == 'linear_density':
self.model = scipy.interpolate.interp1d(list(Counter(X).keys()),
list(Counter(X).values()),
fill_value="extrapolate")
if self.interpolation_type == 'kde':
self.model = mlab.GaussianKDE(X, 'scott')
def test_scalar_covariance_dataset(self):
"""Test a scalar's cov factor."""
np.random.seed(8765678)
n_basesample = 50
multidim_data = [np.random.randn(n_basesample) for i in range(5)]
kde = mlab.GaussianKDE(multidim_data, bw_method=0.5)
assert kde.covariance_factor() == 0.5
def test_callable_singledim_dataset(self):
"""Test the callable's cov factor for a single-dimensional array."""
np.random.seed(8765678)
n_basesample = 50
multidim_data = np.random.randn(n_basesample)
kde = mlab.GaussianKDE(multidim_data, bw_method='silverman')
y_expected = 0.48438841363348911
assert_almost_equal(kde.covariance_factor(), y_expected, 7)
def test_callable_covariance_dataset(self):
"""Test the callable's cov factor for a multi-dimensional array."""
np.random.seed(8765678)
n_basesample = 50
multidim_data = [np.random.randn(n_basesample) for i in range(5)]
def callable_fun(x):
return 0.55
kde = mlab.GaussianKDE(multidim_data, bw_method=callable_fun)
assert kde.covariance_factor() == 0.55
def test_kde_bandwidth_method(self):
np.random.seed(8765678)
n_basesample = 50
xn = np.random.randn(n_basesample)
# Default
gkde = mlab.GaussianKDE(xn)
# Supply a callable
gkde2 = mlab.GaussianKDE(xn, 'scott')
# Supply a scalar
gkde3 = mlab.GaussianKDE(xn, bw_method=gkde.factor)
xs = np.linspace(-7, 7, 51)
kdepdf = gkde.evaluate(xs)
kdepdf2 = gkde2.evaluate(xs)
assert kdepdf.all() == kdepdf2.all()
kdepdf3 = gkde3.evaluate(xs)
assert kdepdf.all() == kdepdf3.all()
def pWH(attributes, players, players_group_name, save_name):
'''
plot the attributes vs height, weight and BMI
players: players of specific group:DataFrame
attributes:list of attributes:list
players_group_name: group name of players:str
save_name: name of saved graph:str
'''
items = attributes
ply = players
pln = players_group_name
itemn = save_name
assert isinstance(items, list)
assert isinstance(ply, pd.DataFrame)
assert isinstance(pln, str)
assert isinstance(itemn, str)
x = [ply.groupby('Hight (cm)').mean(), ply.groupby('Weight').mean()]
l = len(items)
fig, axs = subplots(l, 2, figsize=(12, 6 * l))
for j in range(l):
item = items[j]
for i in range(2):
x1 = x[i].index.values
y1 = x[i][item].values
xv = np.linspace(min(x1), max(x1), 500)
fl = interpolate.interp1d(x1, y1)
xname = x[i].index.name
dx = ply[xname]
normal = mlab.GaussianKDE(dx)(xv)
normal = normal + max(normal) / 2
if l > 1:
ax = axs[j, i]
else:
ax = axs[i]
d = xv[1] - xv[0]
for k in range(499):
ax.add_patch(
patches.Rectangle((xv[k], 0),
d,
fl(xv[k]),
color='b',
linewidth=0,
alpha=normal[k] / max(normal)))
ax.set_xlabel(xname + ' (lb)' * (xname == 'Weight'),
fontsize='x-large')
ax.set_ylabel(item, fontsize='x-large')
if xname == 'Hight (cm)':
xname = 'Height'
ax.set_title(item + ' vs ' + xname + ' for ' + pln,
fontsize='x-large')
ax.grid(True)
ax.set_ylim([0, 100])
ax.set_xlim([min(xv), max(xv)])
fig.savefig('./graph/' + itemn + ' of ' + pln + '.jpg')
def test_evaluate_inv_dim(self):
"""
Invert the dimensions; i.e., for a dataset of dimension 1 [3, 2, 4],
the points should have a dimension of 3 [[3], [2], [4]].
"""
np.random.seed(8765678)
n_basesample = 50
multidim_data = np.random.randn(n_basesample)
kde = mlab.GaussianKDE(multidim_data)
x2 = [[1], [2], [3]]
with pytest.raises(ValueError):
kde.evaluate(x2)
def test_evaluate_diff_dim(self):
"""
Test the evaluate method when the dim's of dataset and points have
different dimensions.
"""
x1 = np.arange(3, 10, 2)
kde = mlab.GaussianKDE(x1)
x2 = np.arange(3, 12, 2)
y_expected = [
0.08797252, 0.11774109, 0.11774109, 0.08797252, 0.0370153
]
y = kde.evaluate(x2)
np.testing.assert_array_almost_equal(y, y_expected, 7)
def test_gaussian_kde_covariance_caching(self):
x1 = np.array([-7, -5, 1, 4, 5], dtype=float)
xs = np.linspace(-10, 10, num=5)
# These expected values are from scipy 0.10, before some changes to
# gaussian_kde. They were not compared with any external reference.
y_expected = [0.02463386, 0.04689208, 0.05395444, 0.05337754,
0.01664475]
# set it to the default bandwidth.
kde2 = mlab.GaussianKDE(x1, 'scott')
y2 = kde2(xs)
np.testing.assert_array_almost_equal(y_expected, y2, decimal=7)
def draw_wage_of_top(top_number, max_of_x):
'''
draw the wage distribution of top players
top_number:number of players:int
max_of_x: limit of axis x
'''
h = top_number
m = max_of_x
assert isinstance(h, int)
assert isinstance(m, int)
item = 'Wage(K)'
dph = defend_player.head(h)
sph = strike_player.head(h)
mph = midfield_player.head(h)
gph = goalkeep_player.head(h)
dpi = dph[item]
spi = sph[item]
mpi = mph[item]
gpi = gph[item]
x = pd.concat([dph, sph, mph, gph])[item]
bins = np.linspace(x.min(), x.max(), 10)
db = bins[1] - bins[0]
x1 = np.linspace(x.min(), x.max(), 100)
normals = norm.pdf(x1, spi.mean(), spi.std()) * h * db
normalm = norm.pdf(x1, mpi.mean(), mpi.std()) * h * db
normald = norm.pdf(x1, dpi.mean(), dpi.std()) * h * db
normalg = norm.pdf(x1, gpi.mean(), gpi.std()) * h * db
kde = mlab.GaussianKDE(x)
plt.hist([sph[item], mph[item], dph[item], mph[item]],
bins=bins,
rwidth=0.8,
edgecolor='k',
stacked=True,
label=['Striker', 'Midfielder', 'Defender', 'GoalKeeper'],
alpha=0.8)
plt.plot(x1, normals, label='Striker', linewidth=3, color='b')
plt.plot(x1, normalm, label='Midfielder', linewidth=3, color='yellow')
plt.plot(x1, normald, label='Defender', linewidth=3, color='lime')
plt.plot(x1, normalg, label='GoalKeeper', linewidth=3, color='red')
plt.grid(True)
plt.xlabel(item, fontsize='x-large')
plt.ylabel('Number of players', fontsize='x-large')
plt.legend(loc='best')
plt.title('Distribution of Wage' + ' of top' + str(h) + ' players',
fontsize='x-large')
plt.xlim([0, m])
def plot_height_weight_BMI(attribute, players):
'''
plot the an attribute vs height, weight and BMI
players: specific players:DataFrame
item:attribute:str
'''
item = attribute
ply = players
x = [
ply.groupby('Hight (cm)').mean(),
ply.groupby('Weight').mean(),
ply.groupby('BMI').mean()
]
assert isinstance(item, str)
assert isinstance(ply, pd.DataFrame)
fig, axs = subplots(1, 3, figsize=(20, 5))
for i in range(3):
x1 = x[i].index.values
y1 = x[i][item].values
xv = np.linspace(min(x1), max(x1), 100)
fl = interpolate.interp1d(x1, y1)
xname = x[i].index.name
dx = ply[xname]
#normal = norm.pdf(xv, dx.mean(), dx.std())
normal = mlab.GaussianKDE(dx)(xv)
normal = normal + max(normal) / 2
ax = axs[i]
d = xv[1] - xv[0]
for j in range(99):
ax.add_patch(
patches.Rectangle((xv[j], 0),
d,
fl(xv[j]),
color='b',
linewidth=0,
alpha=normal[j] / max(normal)))
ax.set_xlabel(xname + ' (lb)' * (xname == 'Weight'),
fontsize='x-large')
ax.set_ylabel(item, fontsize='x-large')
if xname == 'Hight (cm)':
xname = 'Height'
ax.set_title(item + ' vs ' + xname, fontsize='x-large')
ax.grid(True)
ax.set_ylim([0, 100])
ax.set_xlim([min(xv), max(xv)])
fig.savefig('./graph/' + item + '.jpg')
def test_4():
# 读取数据
data = pd.read_excel(
'C:/Users/zhaozehui/PycharmProjects/DataAnalysis/venv/data/hengxiang/test1.xls'
)
datas = data['语文']
print(datas.count())
# 绘图：语文成绩的直方图
plt.hist(
datas, # 绘图数据
bins=np.arange(datas.min(), datas.max(), 3), # 指定直方图的条形数为20个
normed=True, # 设置为频率直方图
color='steelblue', # 指定填充色
align='left',
edgecolor='k', # 指定直方图的边界色
label='直方图' # 为直方图呈现标签
)
plt.title('班级成绩分布直方图')
plt.xlabel('成绩')
plt.ylabel('人数')
# 生成正态曲线的数据
x1 = np.linspace(datas.min(), datas.max(), 1000)
normal = mlab.normpdf(x1, datas.mean(), datas.std())
# 绘制正态分布曲线
line1, = plt.plot(x1, normal, 'r-', linewidth=2)
# 生成核密度曲线的数据
kde = mlab.GaussianKDE(datas)
x2 = np.linspace(datas.min(), datas.max(), 1000)
# 绘制
line2, = plt.plot(x2, kde(x2), 'g-', linewidth=2)
# 去除图形顶部边界和右边界的刻度
plt.tick_params(top='off', right='off')
# 显示图例
plt.legend([line1, line2], ['正态分布曲线', '核密度曲线'], loc='best')
# 显示图形
plt.show()
return
def single_attribute_distribution(attribute, unit=''):
'''
draw distribution of single attribute
attribute(str):the attribute
unit(str)
'''
item = attribute
assert isinstance(item, str)
assert isinstance(unit, str)
x = player_attributes[item]
bins = np.linspace(x.min(), x.max(), 10)
x1 = np.linspace(x.min(), x.max(), 100)
normal = norm.pdf(x1, x.mean(), x.std()) * x.count() * (bins[1] - bins[0])
kde = mlab.GaussianKDE(x)
plt.hist([
defend_player[item], midfield_player[item], strike_player[item],
goalkeep_player[item]
],
bins=bins,
rwidth=0.8,
edgecolor='k',
stacked=True,
label=['Defender', 'Midfielder', 'Striker', 'GoalKeeper'])
plt.plot(x1,
kde(x1) * x.count() * (bins[1] - bins[0]),
linewidth=3,
label='Kernel density')
plt.plot(x1, normal, label='Normal distribution', linewidth=3)
plt.grid(True)
plt.xlabel(item + unit, fontsize='x-large')
plt.ylabel('Number of players', fontsize='x-large')
plt.legend(loc='best')
if item == 'Hight (cm)':
item = 'Height'
plt.title('Distribution of ' + item, fontsize='x-large')
plt.savefig('./graph/' + 'Distribution of ' + item + '.jpg')
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.mlab as mlab
plt.rcParams['font.sans-serif'] = ['Microsoft YaHei']
plt.rcParams['axes.unicode_minus'] = False
titanic = pd.read_csv('birthrate.csv')
titanic.dropna(subset=['2008'], inplace=True)
plt.style.use('ggplot')
plt.hist(titanic['2008'],
bins=np.arange(titanic['2008'].min(), titanic['2008'].max(), 3),
normed=True,
color='steelblue',
edgecolor='k')
plt.title('2008出生直方图和密度图')
plt.xlabel('出生率(‰)')
plt.ylabel('频率(‰)')
kde = mlab.GaussianKDE(titanic['2008'])
x2 = np.linspace(titanic['2008'].min(), titanic['2008'].max(), 1000)
line2 = plt.plot(x2, kde(x2), 'g-', linewidth=2)
plt.tick_params(top='off', right='off')
plt.show()
def test_scott_multidim_dataset(self):
"""Test scott's output for a multi-dimensional array."""
x1 = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
with pytest.raises(np.linalg.LinAlgError):
mlab.GaussianKDE(x1, "scott")
def test_no_data(self):
"""Pass no data into the GaussianKDE class."""
with pytest.raises(ValueError):
mlab.GaussianKDE([])
def _kde_method(X, coords):
if np.all(X[0] == X):
return (X[0] == coords).astype(float)
kde = mlab.GaussianKDE(X, bw_method)
return kde.evaluate(coords)
resid, # 绘图数据
bins=100, # 指定直方图条的个数
normed=True, # 设置为频率直方图
color='steelblue', # 指定填充色
edgecolor='k') # 指定直方图的边界色
# 设置坐标轴标签和标题
plt.title('残差直方图')
plt.ylabel('密度值')
# 生成正态曲线的数据
x1 = np.linspace(resid.min(), resid.max(), 1000)
normal = mlab.normpdf(x1, resid.mean(), resid.std())
# 绘制正态分布曲线
plt.plot(x1, normal, 'r-', linewidth=2, label='正态分布曲线')
# 生成核密度曲线的数据
kde = mlab.GaussianKDE(resid)
x2 = np.linspace(resid.min(), resid.max(), 1000)
# 绘制核密度曲线
plt.plot(x2, kde(x2), 'k-', linewidth=2, label='核密度曲线')
# 去除图形顶部边界和右边界的刻度
plt.tick_params(top='off', right='off')
# 显示图例
plt.legend(loc='best')
# 显示图形
plt.show()
#<><>><><><><><><><><><><><><><><><>
# 残差的正态性检验（PP图和QQ图法）
pp_qq_plot = sm.ProbPlot(resid)
pp_qq_plot.ppplot(line='45')
plt.title('P-P图')
pp_qq_plot.qqplot(line='q')
if k % 2 ==0:
# 25没有数据会有异常
try:
data_dic[k-1] = data_dic[k-1] + g_list
except KeyError as e:
data_dic[int(str(e))]=g_list
# 如果为奇数插入字典
else:
data_dic[k] = g_list
# 格式化输出成茎叶图
print('黄彩思1704010135')
# 遍历字典
x = []
for k,v in data_dic.items():
a = ''
# 循环输出列表数据并拼接成字符串
for i in v:
a = a + ' ' + str(i)
# 格式化输出
# print(str(k).rjust(5), a)
x.append(k)
x.append(57)
plt.hist(last_data, bins=x, edgecolor='k', normed=True)
kde = mlab.GaussianKDE(last_data)
x2 = np.linspace(min(last_data), max(last_data), 1000)
plt.plot(x2, kde(x2), 'g-', linewidth=2)
plt.xlabel('出生率')
plt.ylabel('频率')
plt.show()
ax.set_ylabel('Probability density')
ax.set_xlabel('Similarity(The higher the more similar)')
ax.set_title('Histogram of Similarity')
plt.legend()
plt.show()
fig = plt.figure()
ax = fig.add_subplot(111)
ax.hist(same, bins=bi, normed=True, alpha=0.8, label="Same person")
ax.hist(diff, bins=bi, normed=True, alpha=0.8, label="Different person")
ax.set_ylabel('Probability density')
ax.set_xlabel('Similarity(The higher the more similar)')
ax.set_title('Histogram of Similarity')
x1 = np.linspace(min(diff), max(diff), 1000)
normal = mlab.normpdf(x1, np.mean(diff), np.std(diff))
line1, = plt.plot(x1, normal, 'r-', linewidth=2)
kde = mlab.GaussianKDE(diff)
x2 = np.linspace(min(diff), max(diff), 1000)
line2, = plt.plot(x2, kde(x2), 'g-', linewidth=2)
plt.legend(
[line1, line2],
['normal', 'gussiankde'],
)
x3 = np.linspace(min(same), max(same), 1000)
normal = mlab.normpdf(x3, np.mean(same), np.std(same))
line3, = plt.plot(x3, normal, 'r-', linewidth=2)
kde = mlab.GaussianKDE(same)
x4 = np.linspace(min(same), max(same), 1000)
line4, = plt.plot(x4, kde(x4), 'g-', linewidth=2)
plt.legend([line1, line2], ['normal', 'gussiankde'], loc="best")
plt.show()
def test_evaluate_point_dim_not_one(self):
x1 = np.arange(3, 10, 2)
x2 = [np.arange(3, 10, 2), np.arange(3, 10, 2)]
kde = mlab.GaussianKDE(x1)
with pytest.raises(ValueError):
kde.evaluate(x2)
def test_scott_singledim_dataset(self):
"""Test scott's output a single-dimensional array."""
x1 = np.array([-7, -5, 1, 4, 5])
mygauss = mlab.GaussianKDE(x1, "scott")
y_expected = 0.72477966367769553
assert_almost_equal(mygauss.covariance_factor(), y_expected, 7)
def _kde_method(X, coords):
# fallback gracefully if the vector contains only one value
if np.all(X[0] == X):
return (X[0] == coords).astype(float)
kde = mlab.GaussianKDE(X, bw_method)
return kde.evaluate(coords)
def test_scalar_empty_dataset(self):
"""Test the scalar's cov factor for an empty array."""
with pytest.raises(ValueError):
mlab.GaussianKDE([], bw_method=5)
def test_single_dataset_element(self):
"""Pass a single dataset element into the GaussianKDE class."""
with pytest.raises(ValueError):
mlab.GaussianKDE([42])
def test_silverman_singledim_dataset(self):
"""Test silverman's output for a single dimension list."""
x1 = np.array([-7, -5, 1, 4, 5])
mygauss = mlab.GaussianKDE(x1, "silverman")
y_expected = 0.76770389927475502
assert_almost_equal(mygauss.covariance_factor(), y_expected, 7)
