How to create a matplotlib surface plot from frequencies of categorical data-CodePudding

I have an (example) dataframe that looks like this:

    time                  event type
0   2022-01-22 10:35:00     a
1   2022-01-22 11:37:00     a
2   2022-01-22 22:22:00     b
3   2022-01-22 12:05:00     b
4   2022-01-22 10:09:00     c
5   2022-01-22 10:57:00     a
6   2022-01-22 11:36:00     c
7   2022-01-22 09:45:00     a

I would like to create a 3D surface plot that shows how many event from each type occur per hour. The axis of the plot should be:

 X: hour
 Y: event type
 Y: number of events

I would expect to see on the X axis:9, 10, 11, 12 ,22, on theY axis: a, b, c. As for the Z axis,the values should reflect the number of event per type per hour. E.g. X=10, Y=a, Z=2

I looked at the documentation and various examples, but could not find and answer

CodePudding user response：

This requires two tasks. First, you have to aggregate your data to count unique pairs hour-event type, then create the 3D plot from the aggregated hour-event type-event count data:

from matplotlib import pyplot as plt
from matplotlib.ticker import MaxNLocator
import pandas as pd
import numpy as np

#test data 
np.random.seed(123)
n = 10
start = pd.to_datetime("2021-04-21")
end = pd.to_datetime("2021-04-23")
n_minut = ((end - start).days   1) * 24 * 60
date_range = pd.to_timedelta(np.random.randint(0, n_minut, n), unit="minute")   start
df = pd.DataFrame({"time": date_range, "event type": np.random.choice(list("abcde"), n)})

#count event types per hour
plot_df = df. groupby([df["time"].dt.hour, df["event type"]]).size().reset_index(name="event count")
#transcribe categorical data in column "event type" into integer values
#idx contains the list of event types according to their integer numbers
val, idx = plot_df["event type"].factorize()
plot_df["event_num"] = val

#generate evenly spaced x- and y-values
x_range = np.arange(24)
y_range = np.arange(idx.size)

#and create x-y arrays for the 3D plot
X, Y = np.meshgrid(x_range, y_range)
#and fill z-values with zeros
Z = np.zeros(X.shape)
#or the event count, if exists
Z[plot_df["event_num"], plot_df["time"]] = plot_df["event count"] 

#create figure with a 3D projection axis
fig = plt.figure()
ax = fig.add_subplot(111, projection="3d")

ax.plot_surface(X, Y, Z)

ax.zaxis.set_major_locator(MaxNLocator(integer=True))
ax.set_yticks(y_range, idx)
ax.set_ylabel("event type")
ax.set_xlabel("time (in h)")
ax.set_zlabel("count")

plt.show()

Sample output:

However, it is well-known that matplotlib has sometimes problems plotting surfaces in the correct order of visibility. Depending on your data, you might be better off with a scatter plot:

...
X, Y = np.meshgrid(x_range, y_range)
Z = np.full(X.shape, np.nan)
Z[plot_df["event_num"], plot_df["time"]] = plot_df["event count"] 

fig = plt.figure()
ax = fig.add_subplot(111, projection="3d")

ax.scatter(X, Y, Z)

ax.zaxis.set_major_locator(MaxNLocator(integer=True))
...