假设我有以下时间序列数据框架：

import numpy as np
import pandas as pd
import random
np.random.seed(2019)

# Generate TS
#rng = pd.date_range('2019-01-01', freq='MS', periods=N)
ts = pd.date_range('2000-01-01', '2000-12-31 23:00', freq='H') #.strftime('%d-%m-%Y %H:%M:%S')    # freq='MS'set the frequency of date in months and start from day 1. You can use 'T' for minutes and so on

# number of samples
N = len(ts)

# Create a random dataset
data = {
    #"TS": ts,
    'Appx':  [random.choice(['App1', 'App2', 'App3', None]) for _ in range(N)], # generate categorical data including missing data "None"
    'VM':    [random.choice(['VM1' , 'VM2' ]) for _ in range(N)]
}

df = pd.DataFrame(data, index=ts)
#df.resample('M').mean().plot()
df
#                       Appx    VM
#2000-01-01 00:00:00    App1    VM2
#2000-01-01 01:00:00    None    VM1
#2000-01-01 02:00:00    None    VM2
#2000-01-01 03:00:00    App3    VM2
#2000-01-01 04:00:00    App1    VM1
#...    ... ...
#2000-12-31 19:00:00    App2    VM1
#2000-12-31 20:00:00    App3    VM1
#2000-12-31 21:00:00    App3    VM1
#2000-12-31 22:00:00    App1    VM1
#2000-12-31 23:00:00    App1    VM1
# 8784 rows × 2 columns

查看其他可用资源后：

• Plotting categorical data counts over time
• How to make a line plot from a dataframe with multiple categorical columns in matplotlib
• How to groupby dataframe with categorical variables for making linechart in matplotlib?
• How to plot by category over time
• Plot time series on category level
• Plotting multiple time series after a groupby in pandas
• Pandas: plot multiple time series DataFrame into a single plot
• Plot a line graph with categorical columns for each line
  **问题：**在pandas框架内绘制分类变量的计数记录，包括其缺失值（None或NaN）
  **我的尝试：**我尝试使用以下脚本绘制计数记录，但没有成功：首先，我使用了一个简单的示例，通过绘制每个所需VM随时间推移的缺失值（虚线）内的应用列记录，来描述受here启发的数据和缺失值：

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

day = ([ 1 , 2 , 3, 4, 5 , 6 , 7 , 8 , 9])
App1 = ([0.6 , 0.8 , np.nan, np.nan, 4 , 6 , 6.5 ,7 , 8])
App2 = ([ 1 , 2 , np.nan, np.nan, 0.5 , 7 , 8 , 9 , 10])
App3 = ([ 1.5 , 2.5 , np.nan, np.nan, 3 , 4 , 6 , 8 , 11])

cf = pd.DataFrame({'App1': App1, 'App2': App2, 'App3': App3}, index = day)
cf.index.name = 'day'

fig, ax = plt.subplots()
line, = ax.plot(cf['App1'].fillna(method='ffill'), color='r', ls = '--', lw = 1, label='_nolegend_')
ax.plot(cf['App1'], color='k', lw=1.5, marker = 'v', label='App1',)
line, = ax.plot(cf['App2'].fillna(method='ffill'), color='r', ls = '--', lw = 1, label='_nolegend_')
ax.plot(cf['App2'], color='k', lw=1.5, marker = 's', label='App2')
line, = ax.plot(cf['App3'].fillna(method='ffill'), color='r', ls = '--', lw = 1, label='_nolegend_')
ax.plot(cf['App3'], color='k', lw=1.5, marker = 'o', label='App3')

plt.xlabel('Time Stamp')
plt.ylabel('Record counts')
plt.title('Apps within missing values for VM1')
plt.legend()
plt.show()

到目前为止，我的输出：

但是，当我将其应用于基于此answer生成的时间序列数据时，我得到了错误：

import matplotlib.pyplot as plt

df['Appx'].fillna(value=np.nan, inplace=True)
df['Appx'].astype('category') # or str for string
#df = df.astype(int)

# Filter the DataFrame by a list of string values in the "App1" column
filtered_df  = df[ df["Appx"].isin([np.nan])]
filtered_dff = df[~df["Appx"].isin([np.nan])]

cf = pd.DataFrame({'Appx': filtered_dff["Appx"]}, index = df.index)
#cf.index.name = df.index #'TS'

fig, ax = plt.subplots()
line, = ax.plot(cf['Appx'].fillna(method='ffill'), ls = '--', lw = 1, label='_nolegend_')
ax.plot(cf['Appx'], color=line.get_color(), lw=1.5, marker = 'o')
ax.tick_params(axis='x', labelrotation=45)
plt.xlabel('TS')
plt.ylabel('mm')
plt.legend('best')
plt.show()

TypeError：'value'必须是str或bytes的示例，而不是float
甚至我使用groupby()进一步挖掘：

# reset_index() gives a column for counting, after groupby uses year and category
ctdf = (df.reset_index()
          .groupby(['Appx','VM'], as_index=False)
          .count()
          # rename isn't strictly necessary here, it's just for readability
          .rename(columns={'index':'ct'})
       )
ctdf

#    Appx    VM  ct
#0  App1    VM1 1127
#1  App1    VM2 1084
#2  App2    VM1 1066
#3  App2    VM2 1098
#4  App3    VM1 1084
#5  App3    VM2 1049

df['Appx'].fillna(value=np.nan, inplace=True)
df['Appx'].astype('category') # or str for string
#df = df.astype(int)

# Filter the DataFrame by a list of string values in the "App1" column
filtered_df  = df[ df["Appx"].isin([np.nan])]
#filtered_dff = df[~df["Appx"].isin([np.nan])]

# reset_index() gives a column for counting, after groupby uses year and category
ctdff = (filtered_df
          #.isna()
          .reset_index()
          .groupby(['VM'], as_index=False)
          .count()
          # rename isn't strictly necessary here, it's just for readability
          .rename(columns={'index':'ct'})
       )
ctdff

#   VM   ct   Appx
#0  VM1 1153    0
#1  VM2 1123    0

类似于这个answer我可能感兴趣这样的情节所谓的cat_horizontal_plot：

注意：我对尽可能多地删除或估算解决方案不感兴趣：

• Handling missing categorical values ML
• Applying OneHotEncoding on categorical data with missing values
• replace missing values in categorical data
• Deal with missing categorical data python

在这些极端情况下，我不能显示缺失值：

import seaborn as sns
import matplotlib.pyplot as plt
sns.lineplot(data = df, x = df.index, y = 'Appx', hue = 'Appx', marker='o', alpha=0.2)
plt.legend(bbox_to_anchor=[0.5, 1.02], loc='lower center')
plt.xticks(rotation=45)
plt.show()

grouped = df.groupby(['VM','Appx'])

for key, group in grouped:
    data = group.groupby(lambda x: x.hour).count()
    data['Appx'].plot(label=key , legend=True)