Pandas / Matplotlib - smooth out line graph from multiple DataFrame columns

Question

I would like to smooth out a simple line graph with dummy data in a DataFrame. I'm aware of interpolate but all of the tutorials work with simpler numpy arrays ; i would like each of the lines generated by the columns to be smooth. If anyone has experience, or knows how to do this, would be greatly appreciated.

For reference, here is some code, and the resulting graph.

# amps

df=pd.read_csv("~/data/poli.csv")

#################################

# wp

wave = (df.right + df.left)**2
df['concord'] = wave
print(df)

which outputs:

         issue  right  left  concord
0      end div      1     1        4
1    for trump      1    -1        0
2  aisle cross      1     1        4
3      for blm     -1     1        0
4   help world      1     1        4
5      service      1     1        4
6    community      1     1        4

Then I plot the columns, setting the x axis to 'issues', and the y label to 'score'

plot = df.plot(x='issue',
             linewidth=9,
             alpha=0.36)

plot.set_ylabel('score')
plot.set_xlabel('issues')
plot.legend(bbox_to_anchor=(1, 1), loc='upper left', ncol=1)

concord=df['concord'].values
left=df['left'].values
right=df['right'].values

plt.show()

And the resulting plot is generated.

Smoothing each curve will inherently require additional data points at intermediate x values to be generated. These points will not correspond to either of the categories you currently have. Therefore, you will have to first smooth each column (splines should do?), plot them against a linearly increasing x axis (say, 0 to 30, depending on how many points your smoothing produces), and finally manually adjust the ticks so that each extremum corresponds to an entry in your first column. — Patol75
– Patol75, Commented Apr 27, 2020 at 7:05

Lith · Accepted Answer · 2020-04-27 07:21:18Z

As you said, you need to interpolate the curves to increase resolution. This way, you can keep the known values of your data while still effectivily smoothing the curves. To do that, you could use scipy's one dimensional interpolation: interp1d.

We want to interpolate to a bigger number of x-coordinates than we have, but the problem here is that you have strings instead of numbers as independent variables of your left, right, and concorde "functions". To solve this we can simply map your issue column into a monotonically increasing numerical vector, then interpolate in a new vector with limit values taken from it, and finally keep the x-labels as strings.
The updated code could be:

import numpy as np
import pandas as pd
from scipy import interpolate
import matplotlib.pyplot as plt

d = {
     'issue': ['end div', 'for trump', 'aisle cross', 'for blm', 'help world',
     'service', 'community'],
     'right': [1, 1, 1, -1, 1, 1, 1],
     'left': [1, -1, 1, 1, 1, 1, 1],
     'concord': [4, 0, 4, 0, 4, 4, 4]
     }
df = pd.DataFrame.from_dict(d)

x = df['issue'].values
concord = df['concord'].values
left = df['left'].values
right = df['right'].values

n = len(x)
x_map = np.arange(0,n)
dx = 0.1
x_int = np.arange(0, n - 1, dx)  # vector where we interpolate

# We create the interpolants our three datasets separately:
f_concord = interpolate.interp1d(x_map, concord, 'quadratic')
f_left = interpolate.interp1d(x_map, left, 'quadratic')
f_right = interpolate.interp1d(x_map, right, 'quadratic')

# And interpolate in the resampled x-coordinates:
concord_int = f_concord(x_int)
left_int = f_left(x_int)
right_int = f_right(x_int)

# Finally, plot the interpolated data:
fig, ax = plt.subplots()
ax.plot(x_int, right_int, lw = 9, alpha = 0.36, label = 'right')
ax.plot(x_int, left_int, lw = 9, alpha = 0.36, label = 'left')
ax.plot(x_int, concord_int, lw = 9, alpha = 0.36, label = 'concord')
ax.set_xlabel('issues')
ax.set_ylabel('score')
# Set the correct xticks
ax.set_xticks(x_map)
ax.set_xticklabels(x)
fig.legend(bbox_to_anchor=(0.7, 0.3), loc='upper left', ncol=1)
fig.show()

The result:

Collectives™ on Stack Overflow

Pandas / Matplotlib - smooth out line graph from multiple DataFrame columns

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related