python - Preparing a box diagram for representing accuracy metrics - Stack Overflow

admin管理员组

文章数量:1122846

Following is the portion of data I have:

Parameter	Method	RMSE	MAE	FID
DRCT	CubicInterpolation	80.0887	87.9013	8375.7247
DRCT	KrigingInterpolation	49.3513	70.9927	3823.2367
DRCT	LSTM	84.6987	98.1393	11957
DRCT	LinearInterpolation	36.752	35.6747	134.3787
DRCT	PolynomialInterpolation	48.3893	70.502	6116.5867
DRCT	SplineInterpolation	72.8787	45.774	41.7713
DRCT	VAE	102.5967	90.9553	10570.3807
DRCT	cGAN	0.3267	0.3447	0.0587
DWPT	CubicInterpolation	23.2107	28.142	1088.6787
DWPT	KrigingInterpolation	20.348	25.9827	842.3687
DWPT	LSTM	30.4353	45.2293	2797.9993
DWPT	LinearInterpolation	19.686	3.5553	7.8667
DWPT	PolynomialInterpolation	11.9107	17.8107	517.4547
DWPT	SplineInterpolation	7.2767	4.7913	11.3773
DWPT	VAE	11.948	27.9	1162.7693
DWPT	cGAN	0.2487	0.2907	0.0953

Following is the portion of data I have:

Parameter	Method	RMSE	MAE	FID
DRCT	CubicInterpolation	80.0887	87.9013	8375.7247
DRCT	KrigingInterpolation	49.3513	70.9927	3823.2367
DRCT	LSTM	84.6987	98.1393	11957
DRCT	LinearInterpolation	36.752	35.6747	134.3787
DRCT	PolynomialInterpolation	48.3893	70.502	6116.5867
DRCT	SplineInterpolation	72.8787	45.774	41.7713
DRCT	VAE	102.5967	90.9553	10570.3807
DRCT	cGAN	0.3267	0.3447	0.0587
DWPT	CubicInterpolation	23.2107	28.142	1088.6787
DWPT	KrigingInterpolation	20.348	25.9827	842.3687
DWPT	LSTM	30.4353	45.2293	2797.9993
DWPT	LinearInterpolation	19.686	3.5553	7.8667
DWPT	PolynomialInterpolation	11.9107	17.8107	517.4547
DWPT	SplineInterpolation	7.2767	4.7913	11.3773
DWPT	VAE	11.948	27.9	1162.7693
DWPT	cGAN	0.2487	0.2907	0.0953

I am trying to create a diagram to represent Mean Absolute Error (MAE) comparison for different parameters, i.e., DRCT, DWPT across different methods. The horizontal axis will be used for representing the parameters. There is no vertical axis. Instead, I am looking to create a box for each parameter. The box will represent the MAE. The bottom and top will represent the min and max of MAE for the parameter. MAE from other methods will be represented as points or lines inside the box.

Can someone help me to achieve this? I'm clueless at the moment; regular box plots won't do.

expected figure is attached below. I created it using paint. sorry for the less details.

• python
• matplotlib

Share Improve this question Follow edited Nov 22, 2024 at 9:51 srinivas asked Nov 22, 2024 at 8:33 srinivassrinivas 29911 silver badge1111 bronze badges 2

• It would help to provide a schematic of the expected graph with the values (and computations) that match your example – mozway Commented Nov 22, 2024 at 9:00
• I have updated a representative figure now. – srinivas Commented Nov 22, 2024 at 10:10

Add a comment  | 

2 Answers 2

Sorted by: Reset to default 0

Hopefully, this is what you wanted:

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

data = {
    'Parameter': ['DRCT', 'DRCT', 'DRCT', 'DRCT', 'DRCT', 'DRCT', 'DRCT', 'DRCT', 'DWPT', 'DWPT', 'DWPT', 'DWPT', 'DWPT', 'DWPT', 'DWPT', 'DWPT'],
    'MAE': [87.9013, 70.9927, 98.1393, 35.6747, 70.502, 45.774, 90.9553, 0.3447, 28.142, 25.9827, 45.2293, 3.5553, 17.8107, 4.7913, 27.9, 0.2907]
}

df = pd.DataFrame(data)

plt.figure(figsize=(12, 8))
sns.boxplot(x='MAE', y='Parameter', data=df, orient='h')
sns.stripplot(x='MAE', y='Parameter', data=df, color='orange', jitter=0.2, size=5)

plt.title('Horizontal Boxplot of MAE by Parameter')
plt.xlabel('MAE')
plt.ylabel('Parameter')
plt.show()

AFAIK, there is no direct way to do this, but you can use lower level matplotlib. I am assuming here that all parameters have the same methods in the same order (if not, you'll need to handle the color/legends):

import matplotlib.pyplot as plt
import matplotlib
from matplotlib.collections import LineCollection
from matplotlib.lines import Line2D

required = ['q1', 'med', 'q3', 'whislo', 'whishi']

dicts = {}
vals = []

# compute min/max for boxes
for k, g in df.set_index('Method').groupby('Parameter')['MAE']:
    dicts[k] = (g.agg(['min', 'max'])
                 .rename({'min': 'q1', 'max': 'q3'})
                 .reindex(required)
                 .to_dict()
               )
    vals.append(g)

# box width
w = 0.2
    
# plot boxes
ax = plt.subplot()
ax.bxp(dicts.values(),
       widths=np.repeat(w, len(dicts)),
       showfliers=False)
ax.set_xticklabels(list(dicts))

# plot lines
cmap = matplotlib.colormaps['tab10']
for i, v in enumerate(vals, start=1):
    lc = LineCollection([[(i-w/2, y), (i+w/2, y)] for y in v],
                        colors=cmap.colors[:len(v)]
                       )
    ax.add_collection(lc)

# add legend
colors=cmap.colors[:len(v)]
ax.legend([Line2D([0, 1], [0, 1], color=c) for c in colors],
          v.index
         )

Output:

本文标签： pythonPreparing a box diagram for representing accuracy metricsStack Overflow