Scatter plot

A scatter plot shows the relationship between two variables with points.

Scatter plot

A scatter plot shows the relationship of two numeric variables with points. With the c parameter, points are coloured by class; in classification you immediately see whether the classes separate:

from sklearn.datasets import load_iris
X, y = load_iris(return_X_y=True)
fig, ax = plt.subplots()
sc = ax.scatter(X[:, 2], X[:, 3], c=y, cmap="viridis", s=20)
ax.set_xlabel("taç yaprak uzunluğu")
ax.set_ylabel("taç yaprak genişliği")
fig.colorbar(sc)

In Iris these two features separate the three species almost completely; this explains why the tree found them important.

Truth versus prediction

The clearest way to evaluate a regression model is to plot the true values against the predictions. In a good model the points gather around the y = x diagonal:

ax.scatter(y_test, p, alpha=0.6)
lim = [y_test.min(), y_test.max()]
ax.plot(lim, lim, "r--")   # perfect-prediction line

If the points pile up below the diagonal, the model is systematically under-predicting.

Exercise Plot the classes

Draw columns 3 and 4 of Iris (X[:, 2], X[:, 3]) as a scatter plot with c=y and save it. Then compute each class's mean petal length with [X[y == k, 2].mean() for k in range(3)] and print it with np.round(..., 2).

Show example solution

import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import load_iris

X, y = load_iris(return_X_y=True)

fig, ax = plt.subplots()
ax.scatter(X[:, 2], X[:, 3], c=y)
fig.savefig("iris.png")
print(np.round([X[y == k, 2].mean() for k in range(3)], 2))

Test yourself

Which parameter colours the points by class label?
- s
- alpha
- marker
- c
Show answer

c — c takes the colour values; s sets the size and alpha the transparency.

Draw the points semi-transparent.

ax.scatter(x, y, ____=0.5)

Show answer

ax.scatter(x, y, alpha=0.5)

Match the parameters.

c

Colour values
s

Point size
alpha

Transparency
cmap

Colour map

Next lesson: Histogram

We Are Enhancing Your Experience

Cookie Preferences

Strictly Necessary Cookies

Analytical Cookies