3D Scatterplot with strings in Python

问题:

I tried to do a 3D scatter plot in Python with string categories (i.e. activation functions and solvers for a neural network) on x and y and floating numbers (i.e. accuracy score of NN) on the z axis.

The following example raises the error: ValueError: could not convert string to float: 'str1'

I followed this documentation for 3D plots: https://matplotlib.org/mpl_toolkits/mplot3d/tutorial.html

Any ideas, what might be the problem ? Many thanks in advance!

import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
xs=['str1', 'str2']
print(type(xs))
ys=['str3', 'str4']
print(type(ys))
zs=[1,2]
ax.scatter(xs, ys, zs)

回答1:

You are trying to pass categorical values (strings) as the x and y arguments. This would work for 1d scatter plot but in 3d, you need to define the span/cartesian coordinates. What you mainly want to have is the strings as the x and y-axis ticklabels. To get the desired plot, what you can do is to first plot the numeric values and then re-assign the ticklabels as per your string values.

import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')

xs=['str1', 'str2']
ys=['str3', 'str4']
zs=[1,2]

ax.scatter(range(len(xs)), range(len(xs)), zs)
ax.set(xticks=range(len(xs)), xticklabels=xs,
       yticks=range(len(xs)), yticklabels=xs) 

You can also set the tick labels using

plt.xticks(range(len(xs)), xs)
plt.yticks(range(len(ys)), ys)

The first option using ax however allows you to do the same in one line.