Seaborn's pairplot can be used to quickly visualize multivariable relationships. 1. Basic usage draws a scatter plot of each pair of numerical variables, and displays the distribution of each variable in diagonal lines; 2. Use the hue parameter to color by category and distinguish different categories; 3. Use the diag_kind parameter to set the diagonal chart to 'kde' or 'hist'; 4. Use the height and aspect parameters to adjust the size of the sub-graph; 5. Select specific variables to draw through the vars parameter; it is recommended to use it when there are fewer variables. Large data volumes can be combined with plot_kws to set alpha and s to optimize the display effect. This function is an efficient and intuitive tool in exploratory data analysis.

Seaborn's pairplot is a very practical function to quickly visualize the relationship between multiple variables in the dataset. It plots a scatter plot (a histogram or density plot on the diagonal) for each pair of numerical variables, which is ideal for exploratory data analysis (EDA).

Here is a complete example using seaborn.pairplot using the built-in iris dataset:

1. Basic pairplot example

 import seaborn as sns
import matplotlib.pyplot as plt

# Load the built-in iris dataset iris = sns.load_dataset('iris')

# Create a pairplot
sns.pairplot(iris)
plt.show()

This image will show:

• Non-diagonal: Scatter plots between different features (such as sepal_length vs sepal_width)
• Diagonal: Distribution of each feature (default is a histogram)

2. Color by category (hue parameter)

If your data has classification labels, you can use the hue parameter to color the differences between categories to more clearly see the differences between categories.

 sns.pairplot(iris, hue='species')
plt.show()

In this way, different types of iris (setosa, versicolor, virginica) will be displayed in different colors to facilitate observation of classification boundaries.

3. Customize the diagonal chart type

You can use diag_kind parameter to modify the graph type on the diagonal, such as changing it to a density graph:

 sns.pairplot(iris, hue='species', diag_kind='kde')
plt.show()

It can also be set to 'hist' to display histogram.

4. Control the size and style of the graphics

Although pairplot returns a PairGrid object, you can resize the subgraph by height and aspect :

 sns.pairplot(iris, hue='species', height=2.5, aspect=1.2)
plt.show()

• height : the height of each subgraph
• aspect : aspect ratio

5. Only draw some variables

If you only care about certain columns, you can use vars parameter to select:

 sns.pairplot(iris, 
             hue='species', 
             vars=['sepal_length', 'sepal_width', 'petal_length'])
plt.show()

Tips

• pairplot is suitable for datasets with few features (such as 3 to 6 variables), otherwise the chart will be too dense.
• If the data volume is large (such as tens of thousands of rows), the scatter plot may overlap severely. You can consider adding transparency or adjusting the parameters with plot_kws :

 sns.pairplot(iris, hue='species', plot_kws={'alpha': 0.7, 's': 15})
plt.show()

in:

• alpha : transparency
• s : scatter size

Basically that's it. pairplot is a very "out of the box" tool in EDA. You can see the overall structure and class separability of the data in a few lines of code.

The above is the detailed content of python seaborn pairplot example. For more information, please follow other related articles on the PHP Chinese website!