如何在 ax.scatter 中显示和图例两种颜色？

Question

我想根据标签的不同显示两种颜色。具体来说，当 label = 2 时，点颜色为黑色，其余点遵循 cmap ='Dark2'。我只有一个简单的想法，将点（label = 2）再次重叠，这样的代码，

import matplotlib.pyplot as plt
from sklearn import datasets
from collections import Counter
iris = datasets.load_iris()
X = iris.data
y = iris.target
df = pd.DataFrame(X, columns = iris.feature_names)
fig, ax = plt.subplots(figsize=(12,8))
points = ax.scatter(df.values[:,0], 
    df.values[:,1],
    c = y,
    cmap='Dark2')  #others is follow this cmap

for i in range (len(y)):
    if y[i] == 2:
        ax.scatter(df.values[i,0],df.values[i,1], c = 'k') #when label = 2,points color is black 

handles, _ = points.legend_elements()
labels =sorted([f'{item}: {count}' for item, count in Counter(y).items()])
ax.legend(handles, labels, loc = "lower right",title = 'clusters')  

plt.show()

现在，问题是黑点的颜色仍然遵循以前的颜色图，比如（是灰色而不是黑色）。如何解决这个问题呢？

Answer 1

可能有更优雅的解决方案，但您可以手动向图例添加一个标记：

import pandas as pd
import matplotlib.lines as mlines
import matplotlib.pyplot as plt
from sklearn import datasets
from collections import Counter
iris = datasets.load_iris()

X = iris.data
y = iris.target
df = pd.DataFrame(X, columns = iris.feature_names)

fig, ax = plt.subplots(figsize=(12,8))

df1 = df[y!= 2] 
points = ax.scatter(df1.values[:,0], df1.values[:,1], marker = 'o', c = y[y!=2], cmap='Dark2') 

df2 = df[y== 2] 
points2 = ax.scatter(df2.values[:,0], df2.values[:,1], marker = 'o', color = 'k')

handles, _ = points.legend_elements()
labels = sorted([f'{item}: {count}' for item, count in Counter(y).items()])

one_more = mlines.Line2D([], [], color='k', marker='o', linestyle='None', markersize = handles[0].get_ms())
ax.legend(handles + [one_more], labels, loc = "lower right",title = 'clusters')  
plt.show()

它给出：

Answer 2

您可以首先在 y != 2 的记录上绘制散点图，然后在 y == 2 上绘制散点图，但是这一次，因为您想要所有这些 Black 而不是使用 c，因此您设置了 color='black'这些数据点没有不同的值。

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn import datasets
from collections import Counter

iris = datasets.load_iris()
X = iris.data
y = iris.target
df = pd.DataFrame(X, columns=iris.feature_names)
fig, ax = plt.subplots(figsize=(12, 8))

points = ax.scatter(df.values[np.where(y != 2), 0],
           df.values[np.where(y != 2), 1],
           c=y[np.where(y != 2)], cmap='Dark2')

p2 = ax.scatter(df.values[np.where(y == 2), 0], df.values[np.where(y == 2), 1], color='black')

handles, _ = points.legend_elements()
labels =sorted([f'{item}: {count}' for item, count in Counter(y[np.where( y != 2)]).items()])
ax.legend([*handles, p2], [*labels, f'2: {np.sum(y == 2)}'], loc="lower right",title='clusters')

plt.show()

输出：