python - ax.get_yticks() 在 matplotlib 中对轴进行转换
问题描述
[编辑] 我正在尝试将热图与箱线图混合。这个想法是以离散的间隔划分箱形图并相应地着色。更清楚数据：它是同一分钟内同时使用的令牌的每周数据（因此同一周有多行相关）。样本数据如下：
| | start_hour | week | TokenUsed |
|---:|:--------------------|:-----------|------------:|
| 0 | 2019-12-19 20:20:00 | 2019-12-22 | 8 |
| 1 | 2019-12-19 20:21:00 | 2019-12-22 | 8 |
| 2 | 2019-12-19 20:22:00 | 2019-12-22 | 8 |
| 3 | 2019-12-19 20:23:00 | 2019-12-22 | 8 |
| 4 | 2019-12-19 20:24:00 | 2019-12-22 | 8 |
...
| 43370 | 2020-03-11 11:40:00 | 2020-03-15 | 5 |
| 43371 | 2020-03-11 11:41:00 | 2020-03-15 | 5 |
| 43372 | 2020-03-11 11:42:00 | 2020-03-15 | 5 |
| 43373 | 2020-03-11 11:43:00 | 2020-03-15 | 5 |
| 43374 | 2020-03-11 11:44:00 | 2020-03-15 | 5
然后我使用以下内容生成箱线图：
df= minute_by_minute.set_index("week")
percentile = 95
x,y = [],[]
fig, ax = plt.subplots(figsize = (12,6))
for date in df.index.unique():
y.append(my_boxplot_stats(df.loc[date]["TokenUsed"], percents=[100-percentile,percentile], labels=[date], whis="min/max")[0])
x.append(date)
data_box = pd.DataFrame(y)
coverage = (minute_by_minute["TokenUsed"] < data_box['q3'].max()).sum()/(len(df["TokenUsed"]))
ax.set_title(f"Feature:{feature} : The max {data_box['q3'].max()} at {data_box.set_index('label')['q3'].idxmax()}\n It covers {coverage:.2%} ")
bplot = ax.bxp(y,patch_artist=True)
rectangles = []
ax.yaxis.set_minor_locator(AutoMinorLocator(2))
color_map = minute_by_minute.groupby(by=["week","TokenUsed"])["TokenUsed"].count()
for patch,q1,q3,label in zip(bplot['boxes'], data_box["q1"],data_box["q3"],data_box["label"]):
# print(f"Label {label}")
verts = patch.get_verts().copy()
verts = ax.transAxes.inverted().transform(verts)
print(f"Verts : {verts}")
height = (verts[3] - verts[0]).sum()
width = (verts[1] - verts[0]).sum()
items = len(color_map.loc[label].loc[q1:q3])
height_per_unit = height/items
initial_verts = verts[0].copy()
color_value = (color_map.loc[label]/color_map.loc[label].sum())
count = 0
for value, minutes in color_map.loc[label].loc[q1:q3].to_frame().iterrows():
count += 1
rect = patches.Rectangle((initial_verts[0],initial_verts[1]),
0.5,
height_per_unit,
linewidth=0,
facecolor= sns.light_palette("red",as_cmap=True)(color_value.loc[value]),
zorder=3000)
rectangles.append(rect)
ax.add_patch(rect)
initial_verts = ax.transAxes.inverted().transform(rect.get_verts())
initial_verts = initial_verts[3]
ax.axhline(maxs[feature])
ax.axhline(find_spot, ls="--", c="k",alpha = 0.5, dash_capstyle="round")
plt.xticks(rotation=30)
plt.savefig(f"{feature}_{100-percentile}_{percentile}.png")
plt.show()
这给出了以下结果：
但是，绘制在箱线图上的框的高度由 (height_box_plot/number_of_intervals) 定义。
| TokenUsed | TokenUsed |
|------------:|------------:|
| 5 | 685 |
| 8 | 20 |
| 10 | 1835 |
| 15 | 335 |
| 16 | 595 |
| 21 | 65 |
| 23 | 130 |
| 24 | 270 |
| 26 | 5 |
| 29 | 40 |
| 31 | 130 |
| 32 | 210 |
我需要的是使框与刻度线匹配（对值进行分组）。为此，我尝试使用ax.get_yticks()before 循环，但这显然“转换”了轴并给出了 this 
。我说的变换是什么意思？
如果我在没有 ax.get_yticks() 的情况下得到补丁的顶点 (bplot("boxes")) 并且得到两个不同的值：
# Without ax.get_yticks()
Verts : [[ 0.75 5. ]
[ 1.25 5. ]
[ 1.25 32. ]
[ 0.75 32. ]
[ 0.75 5. ]]
# With ax.get_yticks()
Verts : [[0.01923077 0.07997699]
[0.05769231 0.07997699]
[0.05769231 0.39067894]
[0.01923077 0.39067894]
[0.01923077 0.07997699]]
最小的例子如下！您可以切换ax.get_yticks()以验证箱线图的顶点发生了什么。
#%%
import time
from functools import partial
import matplotlib.patches as patches
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
from matplotlib.cbook import _reshape_2D
from matplotlib.collections import PatchCollection
from matplotlib.ticker import MultipleLocator, AutoMinorLocator
def my_boxplot_stats(X, whis=1.5, bootstrap=None, labels=None,
autorange=False, percents=[25, 75]):
"""Function to generate the statistics of a box plot
"""
def _bootstrap_median(data, N=5000):
# determine 95% confidence intervals of the median
M = len(data)
percentiles = [2.5, 97.5]
bs_index = np.random.randint(M, size=(N, M))
bsData = data[bs_index]
estimate = np.median(bsData, axis=1, overwrite_input=True)
CI = np.percentile(estimate, percentiles)
return CI
def _compute_conf_interval(data, med, iqr, bootstrap):
if bootstrap is not None:
# Do a bootstrap estimate of notch locations.
# get conf. intervals around median
CI = _bootstrap_median(data, N=bootstrap)
notch_min = CI[0]
notch_max = CI[1]
else:
N = len(data)
notch_min = med - 1.57 * iqr / np.sqrt(N)
notch_max = med + 1.57 * iqr / np.sqrt(N)
return notch_min, notch_max
# output is a list of dicts
bxpstats = []
# convert X to a list of lists
X = _reshape_2D(X, "X")
ncols = len(X)
if labels is None:
labels = itertools.repeat(None)
elif len(labels) != ncols:
raise ValueError("Dimensions of labels and X must be compatible")
input_whis = whis
for ii, (x, label) in enumerate(zip(X, labels)):
# empty dict
stats = {}
if label is not None:
stats['label'] = label
# restore whis to the input values in case it got changed in the loop
whis = input_whis
# note tricksyness, append up here and then mutate below
bxpstats.append(stats)
# if empty, bail
if len(x) == 0:
stats['fliers'] = np.array([])
stats['mean'] = np.nan
stats['med'] = np.nan
stats['q1'] = np.nan
stats['q3'] = np.nan
stats['cilo'] = np.nan
stats['cihi'] = np.nan
stats['whislo'] = np.nan
stats['whishi'] = np.nan
stats['med'] = np.nan
continue
# up-convert to an array, just to be safe
x = np.asarray(x)
# arithmetic mean
stats['mean'] = np.mean(x)
# median
med = np.percentile(x, 50)
## Altered line
q1, q3 = np.percentile(x, (percents[0], percents[1]))
# interquartile range
stats['iqr'] = q3 - q1
if stats['iqr'] == 0 and autorange:
whis = 'range'
# conf. interval around median
stats['cilo'], stats['cihi'] = _compute_conf_interval(
x, med, stats['iqr'], bootstrap
)
# lowest/highest non-outliers
if np.isscalar(whis):
if np.isreal(whis):
loval = q1 - whis * stats['iqr']
hival = q3 + whis * stats['iqr']
elif whis in ['range', 'limit', 'limits', 'min/max']:
loval = np.min(x)
hival = np.max(x)
else:
raise ValueError('whis must be a float, valid string, or list '
'of percentiles')
else:
loval = np.percentile(x, whis[0])
hival = np.percentile(x, whis[1])
# get high extreme
wiskhi = np.compress(x <= hival, x)
if len(wiskhi) == 0 or np.max(wiskhi) < q3:
stats['whishi'] = q3
else:
stats['whishi'] = np.max(wiskhi)
# get low extreme
wisklo = np.compress(x >= loval, x)
if len(wisklo) == 0 or np.min(wisklo) > q1:
stats['whislo'] = q1
else:
stats['whislo'] = np.min(wisklo)
# compute a single array of outliers
stats['fliers'] = np.hstack([
np.compress(x < stats['whislo'], x),
np.compress(x > stats['whishi'], x)
])
# add in the remaining stats
stats['q1'], stats['med'], stats['q3'] = q1, med, q3
return bxpstats
#### INPUT DATA #####
np.random.seed(10)
data = pd.DataFrame(pd.date_range("2020-01-01","2020-02-01", freq="1min"),columns=["start_hour"])
data["TokenUsed"] = np.random.normal(15,2, data.shape[0])
data["TokenUsed"] = data["TokenUsed"].astype(int)
minute_by_minute = data.groupby([pd.Grouper(key="start_hour",freq="1W"),"start_hour"]).sum()
minute_by_minute.index.rename(level=0,names="week",inplace=True)
minute_by_minute.reset_index(inplace=True)
percentile = 95
toggle_get_yticks = False
############ BOX PLOT GENERATION
df = minute_by_minute.set_index("week")
x,y = [],[]
fig, ax = plt.subplots(figsize = (12,6))
for date in df.index.unique():
y.append(my_boxplot_stats(df.loc[date]["TokenUsed"], percents=[100-percentile,percentile], labels=[date], whis="min/max")[0])
x.append(date)
data_box = pd.DataFrame(y)
coverage = (df["TokenUsed"] < data_box['q3'].max()).sum()/(len(df["TokenUsed"]))
ax.set_title(f" The max {data_box['q3'].max()} at {data_box.set_index('label')['q3'].idxmax()}\n It covers {coverage:.2%} ")
bplot = ax.bxp(y,patch_artist=True)
rectangles = []
ax.yaxis.set_minor_locator(AutoMinorLocator(2))
if toggle_get_yticks:
ax.get_yticks()
color_map = minute_by_minute.groupby(by=["week","TokenUsed"])["TokenUsed"].count()
######## COLORING THE BOXPLOT
for patch,q1,q3,label in zip(bplot['boxes'], data_box["q1"],data_box["q3"],data_box["label"]):
verts = patch.get_verts().copy()
verts = ax.transAxes.inverted().transform(verts)
print(f"Verts : {verts}")
height = (verts[3] - verts[0]).sum()
width = (verts[1] - verts[0]).sum()
items = len(color_map.loc[label].loc[q1:q3])
height_per_unit = height/items
initial_verts = verts[0].copy()
color_value = (color_map.loc[label]/color_map.loc[label].sum())
count = 0
for value, minutes in color_map.loc[label].loc[q1:q3].to_frame().iterrows():
count += 1
rect = patches.Rectangle((initial_verts[0],initial_verts[1]),
0.5,
height_per_unit,
linewidth=0,
facecolor= sns.light_palette("red",as_cmap=True)(color_value.loc[value]),
zorder=3000)
rectangles.append(rect)
ax.add_patch(rect)
initial_verts = ax.transAxes.inverted().transform(rect.get_verts())
initial_verts = initial_verts[3]
plt.xticks(rotation=30)
plt.show()
# %%
有没有办法阻止或扭转这种转变？
解决方案
让我们看看我是否理解正确。
我正在计算每周令牌的数量，并使用可自定义的 binsize 进行分类。我不确定你的意思是如何完成分箱。为了绘制热图，我正在创建一个使用框轮廓裁剪的图像。
#### INPUT DATA #####
np.random.seed(10)
data = pd.DataFrame(pd.date_range("2020-01-01","2020-02-01", freq="1min"),columns=["start_hour"])
data["TokenUsed"] = np.random.normal(15,2, data.shape[0])
data["TokenUsed"] = data["TokenUsed"].astype(int)
minute_by_minute = data.groupby([pd.Grouper(key="start_hour",freq="1W"),"start_hour"]).sum()
minute_by_minute.index.rename(level=0,names="week",inplace=True)
minute_by_minute.reset_index(inplace=True)
############ BOX PLOT GENERATION
df = minute_by_minute.set_index("week")
x,y = [],[]
for date in df.index.unique():
y.append(my_boxplot_stats(df.loc[date]["TokenUsed"], percents=[100-percentile,percentile], labels=[date], whis="min/max")[0])
x.append(date)
data_box = pd.DataFrame(y)
coverage = (df["TokenUsed"] < data_box['q3'].max()).sum()/(len(df["TokenUsed"]))
ax.set_title(f" The max {data_box['q3'].max()} at {data_box.set_index('label')['q3'].idxmax()}\n It covers {coverage:.2%} ")
fig, ax = plt.subplots()
bplot = ax.bxp(y,patch_artist=True, widths=widths,
medianprops=dict(color='k', ls='--', lw=2))
ax.yaxis.set_minor_locator(AutoMinorLocator(2))
####### EDITS START HERE
binSize = 2
cmap = sns.light_palette("red",as_cmap=True)
widths = 0.8
xlim = ax.get_xlim()
ylim = ax.get_ylim()
# get counts of tokens
token_count = minute_by_minute.groupby(by=["week","TokenUsed"])["TokenUsed"].count()
minToken, maxToken = minute_by_minute['TokenUsed'].agg(['min','max'])
minToken = int(np.floor(minToken/binSize)*binSize)
maxToken = int(np.ceil(maxToken/binSize)*binSize)
binned_tokens = minute_by_minute.groupby(['week',pd.cut(minute_by_minute['TokenUsed'], bins=range(minToken,maxToken+1,binSize))])['TokenUsed'].count()
vmin = binned_tokens.min()
vmax = binned_tokens.max()
for i,((w,gr),box) in enumerate(zip(binned_tokens.groupby(level=0),bplot['boxes'])):
y = np.vstack([gr.values,gr.values])
im = ax.imshow(y.T, aspect='auto', extent=(i+1-widths/2,i+1+widths/2,minToken,maxToken), cmap=cmap, vmin=vmin, vmax=vmax)
box.set_facecolor('none')
im.set_clip_path(box)
ax.set_xlim(xlim)
ax.set_ylim(ylim)
fig.colorbar(im, label='# tokens per bin')
fig.autofmt_xdate()
推荐阅读
- python - 将用 python 训练的 XGBoost 模型移植到另一个用 C/C++ 编写的系统
- javascript - 如何为列表的每条记录添加 data-test-id
- node.js - 使用猫鼬查询计算交易项目总数和交易项目
- javascript - 匹配心脏表情符号字符的 JavaScript 正则表达式行为怪异
- r - 绘制矩阵 ggplot2 的行
- java - 只收集第一个索引
- python - 打印包含用户定义对象的列表返回地址并在 python 中实现复数的根
- javascript - Keccak -f Round constants 十六进制转二进制不是一个位
- python - ValueError: x 和 y 必须具有相同的第一维，但具有形状 (10,) 和 (1, 10)
- apache-spark - 在 Windows 操作系统上安装 Apache-Airflow 时出错