如何在 Python 中创建分类气泡图？

Posted 2023-02-18

【中文标题】如何在 Python 中创建分类气泡图？

【英文标题】：How to create a categorical bubble plot in Python?

【发布时间】：2018-12-18 17:12:38

【问题描述】：

寻求帮助以创建类似于此链接中的情节，仅使用 python 库。Catagorical Bubble Chart using ggplot2 in R：查看投票最多的回复。

这里我借用了链接中的数据：

    df = pd.DataFrame('Var1':['Does.Not.apply',
                                'Not.specified',
                    'Active.Learning..general.',
                       'Problem.based.Learning',
                               'Project.Method',
                          'Case.based.Learning',
                                'Peer.Learning',
                                        'Other',
                               'Does.Not.apply',
                                'Not.specified',
                               'Does.Not.apply',
                    'Active.Learning..general.',
                               'Does.Not.apply',
                       'Problem.based.Learning',
                               'Does.Not.apply',
                               'Project.Method',
                               'Does.Not.apply',
                          'Case.based.Learning',
                               'Does.Not.apply',
                                'Peer.Learning',
                               'Does.Not.apply',
                                       'Other'],
                       'Var2':['Does.Not.apply',
                               'Does.Not.apply',
                               'Does.Not.apply',
                               'Does.Not.apply',
                               'Does.Not.apply',
                               'Does.Not.apply',
                               'Does.Not.apply',
                               'Does.Not.apply',
                                'Not.specified',
                                'Not.specified',
                    'Active.Learning..general.',
                    'Active.Learning..general.',
                       'Problem.based.Learning',
                       'Problem.based.Learning',
                               'Project.Method',
                               'Project.Method',
                          'Case.based.Learning',
                          'Case.based.Learning',
                                'Peer.Learning',
                                'Peer.Learning',
                                        'Other',
                                        'Other'],
                        'Count' : [53,15,1,2,4,22,6,1,15,15,1,1,2,2,4,4,22,22,6,6,1,1])

【参考方案1】：

Plotnine是基于r的ggplot2的图形python实现语法。

代码与您的 R 链接中的代码几乎相同。

import math
import pandas as pd
from plotnine import *

df = pd.DataFrame(<dataframe data here>)

df['dotsize'] = df.apply(lambda row: math.sqrt(float(row.Count) / math.pi)*7.5, axis=1)

(ggplot(df, aes('Var1', 'Var2')) + \
       geom_point(aes(size='dotsize'),fill='white') + \
       geom_text(aes(label='Count'),size=8) + \
       scale_size_identity() + \
       theme(panel_grid_major=element_line(linetype='dashed',color='black'),
             axis_text_x=element_text(angle=90,hjust=1,vjust=0))
).save('mygraph.png')

【参考方案2】：

Python 原生的matplotlib 当然可以创建这种图。它只是一个具有可变标记大小的分类散点图。使用您的玩具数据集：

import numpy as np
import matplotlib.pyplot as plt
import pandas as pd

#create markersize column from values to better see the difference
#you probably want to edit this function depending on min, max, and range of values
df["markersize"] = np.square(df.Count) + 10
fig = plt.figure()
#plot categorical scatter plot
plt.scatter(df.Var1, df.Var2, s = df.markersize, edgecolors = "red", c = "white", zorder = 2)
#plot grid behind markers
plt.grid(ls = "--", zorder = 1)
#take care of long labels
fig.autofmt_xdate()
plt.tight_layout()
plt.show()

输出：

关于散点图的标记大小函数的定义，you might want to read this answer.

【参考方案3】：

解决此问题的另一种方法是 plot an annotation 在每个分类点处使用值和围绕它的圆圈：

import numpy as np
import matplotlib.pyplot as plt
import pandas as pd

#create padding column from values for circles that are neither too small nor too large
df["padd"] = 2.5 * (df.Count - df.Count.min()) / (df.Count.max() - df.Count.min()) + 0.5
fig = plt.figure()
#prepare the axes for the plot - you can also order your categories at this step
s = plt.scatter(sorted(df.Var1.unique()), sorted(df.Var2.unique(), reverse = True), s = 0)
s.remove
#plot data row-wise as text with circle radius according to Count
for row in df.itertuples():
    bbox_props = dict(boxstyle = "circle, pad = ".format(row.padd), fc = "w", ec = "r", lw = 2)
    plt.annotate(str(row.Count), xy = (row.Var1, row.Var2), bbox = bbox_props, ha="center", va="center", zorder = 2, clip_on = True)

#plot grid behind markers
plt.grid(ls = "--", zorder = 1)
#take care of long labels
fig.autofmt_xdate()
plt.tight_layout()
plt.show()

样本输出：

感谢 DavidG，他向我展示了 in this answer 如何防止注释打印在图表之外。