每个细胞具有不同色标的seaborn的混淆矩阵

发布于 2025-01-11 12:43:33 字数 1062 浏览 0 评论 0原文

我想用True/False Positive/Negative 创建混淆矩阵图。

目标是使数字在 [0.0, 1.0] 范围内，对于 True Negative/Positive，色标为 red->green< /code>，假阴性/阳性则为绿色->红色。我在seaborn文档中没有找到任何可以这样做的内容，这可能吗？它可以快速浏览数据，任何红色的东西都会有问题。

目标是左上和右下单元格具有与左下和右上不同的颜色图。这可能吗？

示例代码：

import numpy as np;
import seaborn as sns; 
sns.set_theme()
uniform_data = np.ndarray(shape=(2,2))
uniform_data[0][0] = 0.1
uniform_data[0][1] = 0.9
uniform_data[1][0] = 0.05
uniform_data[1][1] = 0.95
ax = sns.heatmap(uniform_data, annot=True, cmap=sns.color_palette("vlag", as_cmap=True))

图像中的结果：

但我想要的是左上角为红色（因为 0.1 在这里是一个不好的值），右下角为蓝色（因为这里 0.95 是好的值）

所以有些东西像这样（用油漆快速完成，所以不要介意错误，但情感是存在的）：

原文

I want to create a plot of a confusion matrix with True/False Positive/Negative.

The goal is to have the numbers in the range [0.0, 1.0], and for True Negative/Positive, the color scale be red->green, and for the False Negative/Positive for it to be green->red. I have not found anything in the seaborn documentation to do so, is it even possible? It would allow a quick glance over the data, and anything red would be a problem.

The goal would be left top and bottom right cells have a different color map than left bottom and right top. Is this possible?

Example code:

import numpy as np;
import seaborn as sns; 
sns.set_theme()
uniform_data = np.ndarray(shape=(2,2))
uniform_data[0][0] = 0.1
uniform_data[0][1] = 0.9
uniform_data[1][0] = 0.05
uniform_data[1][1] = 0.95
ax = sns.heatmap(uniform_data, annot=True, cmap=sns.color_palette("vlag", as_cmap=True))

results in the image:

But what I would want is to be top left red (as 0.1 is a bad value here), and bottom right blue (as 0.95 is good here)

So something like this (done quickly with paint, so don't mind the errors, but the sentiment is there):

分享到QQ

分享到微博

如果你对这篇内容有疑问，欢迎到本站社区发帖提问参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。

发布评论

需要登录才能够评论，你可以免费注册一个本站的账号。

无尽的现实 2025-01-18 12:43:33

看来您想反转对角线上元素的颜色。 sns.heatmap 的 data 参数决定颜色，而 annot= 参数可以为注释设置不同的内容。修改数据并保留注释的原始内容：

import numpy as np
import seaborn as sns

sns.set_theme()
uniform_data = np.ndarray(shape=(2, 2))
uniform_data[0][0] = 0.1
uniform_data[0][1] = 0.9
uniform_data[1][0] = 0.05
uniform_data[1][1] = 0.95
ax = sns.heatmap(np.abs(np.eye(2) - uniform_data), annot=uniform_data, cmap="vlag", vmin=0, vmax=1)

It seems you want to invert the colors of the elements on the diagonal. The data argument of sns.heatmap decides on the colors, while the annot= argument can set something different for the annotation. Modifying the data and keeping the original for the annotation:

import numpy as np
import seaborn as sns

sns.set_theme()
uniform_data = np.ndarray(shape=(2, 2))
uniform_data[0][0] = 0.1
uniform_data[0][1] = 0.9
uniform_data[1][0] = 0.05
uniform_data[1][1] = 0.95
ax = sns.heatmap(np.abs(np.eye(2) - uniform_data), annot=uniform_data, cmap="vlag", vmin=0, vmax=1)