背景

分類データを扱う際、PCAやNMDSで次元削除した後、matplotlib/seabornで散布図を書くのですが、
「分類ごとに楕円を描いてほしい」と思ったが、なかなか見つからなかったので、メモです。

ドキュメント

コード

# ライブラリのインポート
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.patches import Ellipse
import matplotlib.transforms as transforms

# 楕円を描く関数
# https://github.com/matplotlib/matplotlib/blob/main/examples/statistics/confidence_ellipse.py
def confidence_ellipse(x, y, ax, n_std=3.0, facecolor='none', **kwargs):
    """
    Create a plot of the covariance confidence ellipse of *x* and *y*.

    Parameters
    ----------
    x, y : array-like, shape (n, )
        Input data.

    ax : matplotlib.axes.Axes
        The axes object to draw the ellipse into.

    n_std : float
        The number of standard deviations to determine the ellipse's radiuses.

    **kwargs
        Forwarded to `~matplotlib.patches.Ellipse`

    Returns
    -------
    matplotlib.patches.Ellipse
    """
    if x.size != y.size:
        raise ValueError("x and y must be the same size")

    cov = np.cov(x, y)
    pearson = cov[0, 1]/np.sqrt(cov[0, 0] * cov[1, 1])
    # Using a special case to obtain the eigenvalues of this
    # two-dimensional dataset.
    ell_radius_x = np.sqrt(1 + pearson)
    ell_radius_y = np.sqrt(1 - pearson)
    ellipse = Ellipse((0, 0), width=ell_radius_x * 2, height=ell_radius_y * 2,
                      facecolor=facecolor, **kwargs)

    # Calculating the standard deviation of x from
    # the squareroot of the variance and multiplying
    # with the given number of standard deviations.
    scale_x = np.sqrt(cov[0, 0]) * n_std
    mean_x = np.mean(x)

    # calculating the standard deviation of y ...
    scale_y = np.sqrt(cov[1, 1]) * n_std
    mean_y = np.mean(y)

    transf = transforms.Affine2D() \
        .rotate_deg(45) \
        .scale(scale_x, scale_y) \
        .translate(mean_x, mean_y)

    ellipse.set_transform(transf + ax.transData)
    return ax.add_patch(ellipse)

使用してみた

iris

from sklearn.datasets import load_iris
import seaborn as sns
import pandas as pd
from sklearn.decomposition import PCA

# データの読み込み
iris = load_iris()
df = pd.DataFrame(iris.data, columns=iris.feature_names)
# 標準化
df = df.iloc[:, :].apply(lambda x: (x-x.mean())/x.std(), axis=0)

# PCA
_n_components=2
pca = PCA(n_components=_n_components)
pca.fit(df)
feature = pca.transform(df)
feature_df = pd.DataFrame(feature, columns=["PC{}".format(x + 1) for x in range(_n_components)])

# 目的変数の追加
feature_df["label"] = iris.target
feature_df["label"] = feature_df["label"].map({0: iris.target_names[0], 1: iris.target_names[1], 2: iris.target_names[2]})

# seabornで散布図
colors = dict({"setosa": "blue", 
                "virginica": "orange", 
                "versicolor": "green"})

fig, ax = plt.subplots()
sns.scatterplot(data=feature_df, x="PC1", y="PC2", hue="label", palette=colors, ax=ax)
# 楕円の追加
for name, color in colors.items():
    _x = feature_df[feature_df["label"] == name]["PC1"].values
    _y = feature_df[feature_df["label"] == name]["PC2"].values
    confidence_ellipse(_x, _y, ax, edgecolor=color)
plt.show()

参考にさせていただきました