Seaborn 线图 - 基于峰值的数据

Question

I would like to colour my the lines of a seaborn lineplot, based on where the peak lies. This is my current plot

As seen, some days the peak is in bin 4.72, whilst in others its 5.24 and 5.83. I would like to colour based on these peaks. So for the plot below, it would have 3 colours, whilst maintaining the dates in the legend.

This is my panda's dataframe, called select_bins

               2.79  3.1  3.44  3.82  4.25  4.72  5.24  5.83  6.47  7.19  7.99  8.88
date                                                                           
20180527     1   28   101   270   694  1253  1134   528   106    10     0     0
20180603     0    0     0     3    12    26    82    45     5     0     0     0
20180611     2    7    34   137   317   341   410   179    48    10     1     0
20180617     2    6    13    52   130   133   161    74    23     4     0     0
20180625     0    2     1     9    14    34    47    53     9     0     0     0
20180626     5    1     1     5    18    50    72   101    28     2     0     0
20180628     2    0     0     2    21    41    87    78    16     0     0     0
20180705     1    1     0     2    18    32    63    61    27     7     0     0
20180709     2    0     3     6    31    56   107   139    52    12     1     0

This is the code to plot. As you can see, I transpose the select_bins dataframe to plot

ax = sns.lineplot(data = select_bins.T, dashes=False, palette = 'rocket')
plt.show()

Answer 1

You could group your data and assign individual color palettes:

import seaborn as sns
import pandas as pd
from matplotlib import pyplot as plt
import numpy as np

select_bins = pd.read_csv("test.txt", sep="\s{2,}", engine="python", index_col="date")

#identify unique peaks
col_bin = select_bins.idxmax(axis=1)
unique_val = np.unique(col_bin)

#provide information for palettes
palettes = ["Reds", "Blues", "Greys"]

fig, ax = plt.subplots()

#plot subgroups with their palettes, providing still individual colors within the palette for each line
for uv, pal in zip (unique_val, palettes):
    sns.lineplot(data = select_bins[col_bin==uv].T, dashes=False, palette = pal, ax=ax)

plt.show()

Sample output:

Alternatively, you could use different line styles for the groups but for this you have to reshape your data first from wide to long form . And since we have to convert the date to a string anyhow, why not converting the x-values to numbers for a more realistic representation of the curves?

import seaborn as sns
import pandas as pd
from matplotlib import pyplot as plt

select_bins = pd.read_csv("test.txt", sep="\s{2,}", engine="python", index_col="date")
#identify columns to plot
cols=select_bins.columns
#identify peaks
select_bins["col_bin"] = select_bins.idxmax(axis=1)

#reshape data for plotting
plot_df = select_bins.reset_index().melt(id_vars=["date", "col_bin"], value_vars=cols)
plot_df = plot_df.astype({"date": str, "variable": float})

fig, ax = plt.subplots(figsize=(10, 6))
sns.lineplot(data = plot_df, x="variable", y="value", hue="date", style="col_bin", palette = "rocket", ax=ax)
plt.xticks(ticks=[float(x) for x in cols], labels=cols)

plt.show()

Sample output:

Answer 2

If you create another variables in your dataframe, encoding a value for the peaks [1,2,3] for each line and then set this to hue :

sns.lineplot(data = select_bins.T, dashes=False, palette = 'rocket', hue="peak_encoding")