代码之家 › 专栏 › 技术社区 › user517696

从数据框中绘制sankey图

sankey-diagram plotly pandas

user517696 · 技术社区 · 6 年前

我有一个数据框架:

Vendor Name                 Category                    Count
AKJ Education               Books                       846888
AKJ Education               Computers & Tablets         1045
Amazon                      Books                       1294423
Amazon                      Computers & Tablets         42165
Amazon                      Other                       415
Flipkart                    Books                       1023

我正试图使用上面的数据框架绘制一个sankey图,其来源是 厂商名称 目标是类别流或宽度是伯爵 . 我试图用得很巧妙,但没有成功。有没有人能用plotlyly来制作一个sankey图?

谢谢

2 回复 | 直到 5 年前

vestland 5 年前

这篇文章的答案 How to define the structure of a sankey diagram using a dataframe? 将向您显示,将您的sankey数据源强制转换为一个数据帧可能会很快导致混淆。将节点与链接分离会更好,因为它们的构造不同。

因此,您的节点数据框应该如下所示:

ID               Label    Color
0        AKJ Education  #4994CE
1               Amazon  #8A5988
2             Flipkart  #449E9E
3                Books  #7FC241
4  Computers & tablets  #D3D3D3
5                Other  #4994CE

您的链接数据框应该如下所示:

Source  Target      Value      Link Color
0       3          846888      rgba(127, 194, 65, 0.2)
0       4            1045      rgba(127, 194, 65, 0.2)
1       3         1294423      rgba(211, 211, 211, 0.5)
1       4           42165      rgba(211, 211, 211, 0.5)
1       5             415      rgba(211, 211, 211, 0.5)
2       5               1      rgba(253, 227, 212, 1)

现在,如果您使用类似于苏格兰公投图的设置 plot.ly ,您将能够构建:

由于数字之间的巨大差异,这个特殊的图表看起来有点奇怪。为了说明问题,我把你所有的号码都换成了 1 :

以下是简单复制和粘贴到Jupyter笔记本的全部内容:

# imports
import pandas as pd
import numpy as np
import plotly.graph_objs as go
from plotly.offline import download_plotlyjs, init_notebook_mode, plot, iplot
init_notebook_mode(connected=True)

# Nodes & links
nodes = [['ID', 'Label', 'Color'],
        [0,'AKJ Education','#4994CE'],
        [1,'Amazon','#8A5988'],
        [2,'Flipkart','#449E9E'],
        [3,'Books','#7FC241'],
        [4,'Computers & tablets','#D3D3D3'],
        [5,'Other','#4994CE'],]

# links with your data
links = [['Source','Target','Value','Link Color'],

        # AKJ
        [0,3,1,'rgba(127, 194, 65, 0.2)'],
        [0,4,1,'rgba(127, 194, 65, 0.2)'],

        # Amazon
        [1,3,1,'rgba(211, 211, 211, 0.5)'],
        [1,4,1,'rgba(211, 211, 211, 0.5)'],
        [1,5,1,'rgba(211, 211, 211, 0.5)'],

        # Flipkart
        [2,5,1,'rgba(253, 227, 212, 1)'],
        [2,3,1,'rgba(253, 227, 212, 1)'],]

# links with some data for illustrative purposes ################
#links = [
#    ['Source','Target','Value','Link Color'],
#    
#    # AKJ
#    [0,3,846888,'rgba(127, 194, 65, 0.2)'],
#    [0,4,1045,'rgba(127, 194, 65, 0.2)'],
#    
#    # Amazon
#    [1,3,1294423,'rgba(211, 211, 211, 0.5)'],
#    [1,4,42165,'rgba(211, 211, 211, 0.5)'],
#    [1,5,415,'rgba(211, 211, 211, 0.5)'],
#    
#    # Flipkart
#    [2,5,1,'rgba(253, 227, 212, 1)'],]
#################################################################


# Retrieve headers and build dataframes
nodes_headers = nodes.pop(0)
links_headers = links.pop(0)
df_nodes = pd.DataFrame(nodes, columns = nodes_headers)
df_links = pd.DataFrame(links, columns = links_headers)

# Sankey plot setup
data_trace = dict(
    type='sankey',
    domain = dict(
      x =  [0,1],
      y =  [0,1]
    ),
    orientation = "h",
    valueformat = ".0f",
    node = dict(
      pad = 10,
    # thickness = 30,
      line = dict(
        color = "black",
        width = 0
      ),
      label =  df_nodes['Label'].dropna(axis=0, how='any'),
      color = df_nodes['Color']
    ),
    link = dict(
      source = df_links['Source'].dropna(axis=0, how='any'),
      target = df_links['Target'].dropna(axis=0, how='any'),
      value = df_links['Value'].dropna(axis=0, how='any'),
      color = df_links['Link Color'].dropna(axis=0, how='any'),
  )
)

layout = dict(
        title = "Draw Sankey Diagram from dataframes",
    height = 772,
    font = dict(
      size = 10),)

fig = dict(data=[data_trace], layout=layout)
iplot(fig, validate=False)

Trupti 6 年前

我用过 ggalluvial 和 alluvial 绘制sankey图的库。尤其是, G冲积层 最适合使用,因为它有多种选择和足够的文献。