Matplotlib：用不同的线/标记/颜色绘制数据的插值-d部分

Question

我希望能够用不同的 FMT 绘制数据中插值 d（红色圆圈）的部分。

我目前的想法是绘制 2 条线，原始且仅数据的插值 d 部分，使用不同的 FMT 来区分插值部分，使其看起来像一条奇异线。但我无法提取已被 interpolate-d 的数据，我不确定我的想法目前是否可行，或者是否有更简单的替代方案来完成它。

原始数据

                         Phoenix  Chicago
datetime                             
2012-10-02 10:00:00     30.0     63.0
2012-10-02 11:00:00     30.0     63.0
2012-10-02 12:00:00     31.0     62.0
2012-10-02 13:00:00     31.0     62.0
2012-10-02 14:00:00     32.0     62.0
2012-10-02 15:00:00     17.0     51.0
2012-10-02 16:00:00     28.0      NaN
2012-10-02 17:00:00      NaN      NaN
2012-10-02 18:00:00      9.0     42.0
2012-10-02 19:00:00      NaN      NaN
2012-10-02 20:00:00     13.0      NaN
2012-10-02 21:00:00      NaN      NaN
2012-10-02 22:00:00      7.0     48.0
2012-10-02 23:00:00      6.0     48.0

plt.plot(df[['Phoenix','Chicago']], marker = 'o')
plt.show()

插值后

                     Phoenix  Chicago
datetime                             
2012-10-02 10:00:00     30.0     63.0
2012-10-02 11:00:00     30.0     63.0
2012-10-02 12:00:00     31.0     62.0
2012-10-02 13:00:00     31.0     62.0
2012-10-02 14:00:00     32.0     62.0
2012-10-02 15:00:00     17.0     51.0
2012-10-02 16:00:00     28.0     48.0
2012-10-02 17:00:00     18.5     45.0
2012-10-02 18:00:00      9.0     42.0
2012-10-02 19:00:00     11.0     43.5
2012-10-02 20:00:00     13.0     45.0
2012-10-02 21:00:00     10.0     46.5
2012-10-02 22:00:00      7.0     48.0
2012-10-02 23:00:00      6.0     48.0

df.interpolate(inplace = True)
plt.plot(df[['Phoenix','Chicago']], marker = 'o')
plt.show()

我想要达到的目标的粗略想法

Answer 1

由于 matplotlib 忽略了缺失数据（这通常被认为是一个问题），我们可以在这里使用这个事实，简单地将原始数据绘制在插值数据上：

from matplotlib import pyplot as plt
import pandas as pd

#recreate your data
from io import StringIO
data = """datetime                    Phoenix  Chicago                           
2012-10-02 10:00:00     30.0     63.0
2012-10-02 11:00:00     30.0     63.0
2012-10-02 12:00:00     31.0     62.0
2012-10-02 13:00:00     31.0     62.0
2012-10-02 14:00:00     32.0     62.0
2012-10-02 15:00:00     17.0     51.0
2012-10-02 16:00:00     28.0      NaN
2012-10-02 17:00:00      NaN      NaN
2012-10-02 18:00:00      9.0     42.0
2012-10-02 19:00:00      NaN      NaN
2012-10-02 20:00:00     13.0      NaN
2012-10-02 21:00:00      NaN      NaN
2012-10-02 22:00:00      7.0     48.0
2012-10-02 23:00:00      6.0     48.0"""
df = pd.read_csv(StringIO(data), sep = "\s{2,}", engine="python", parse_dates=["datetime"], na_values = "NaN")
df = df.set_index("datetime")

#interpolate and plot the interpolated data with transparency and dashed lines
dfint = df.interpolate()
plt.plot(dfint[['Phoenix','Chicago']], marker = 'o', alpha=0.5, ls="--")
#reset color cycle, so that the colors match
plt.gca().set_prop_cycle(None)
#plot again with missing data
plt.plot(df[['Phoenix','Chicago']], marker = 'o')
plt.show()

示例输出：

Answer 2

你可以使用.interpolate()方法：

plt.plot(df[['Phoenix','Chicago']].interpolate(), marker = 'o')

编辑（编辑问题后）：

如果您想同时绘制插值版本和未插值版本，您可以简单地在其上绘制未插值版本：

plt.plot(df[['Phoenix','Chicago']].interpolate(), marker = 'x', ls="--")
plt.plot(df[['Phoenix','Chicago']].interpolate(), marker = 'o')

附言。复制数据：

df = pd.DataFrame({"datetime": pd.period_range(start='2012-10-02 10:00:00', end='2012-10-02 23:00:00', freq='H').to_timestamp(),
                   "Phoenix": [30., 30., 31., 31., 32., 17., 28., np.nan, 9., np.nan, 13., np.nan, 7., 6.],
                   "Chicago": [63., 63., 62., 62., 62., 51., np.nan, np.nan, 42., np.nan, np.nan, np.nan, 48., 48.]})\
                       .set_index("datetime")