Question

我有一个如下所示的数据框：

Direction ID    RY Value Part_of_prog
      North  1  2016    29            0
      North  2  2016    25            1
      North  3  2016    25            0
      North  4  2016    22            1
      North  4  2017    22            1
      North  4  2017    24            1
      North  4  2017    25            0
      North  4  2017    26            0
      North  5  2018    26            0
      North  5  2018    26            1
      North  5  2018    20            1
      North  5  2018    22            0
      South  6  2018    22            0
      South  6  2018    22            1
      South  6  2018    24            1
      South  6  2018    24            0
      South  7  2017    24            0
      South  7  2017    24            1
      South  7  2017    19            1
      South  7  2017    18            0
      South  7  2016    13            0
      South  7  2016    13            1
      South  7  2016    14            1
      South  7  2016    19            0
      East  1  2016    29             0
      East  2  2016    15             1
      East  3  2016    25             0
      East  4  2016    22             1
      East  4  2017    22             1
      East  4  2017    14             1
      East  4  2017    25             0
      East  4  2017    26             0
      East  5  2018    16             0
      East  5  2018    26             1
      East  5  2018    10             1
      East  5  2018    22             0

我正在使用下面的代码使用seaborn在Facet网格中绘制条形图：

g = sns.FacetGrid(df,col='Direction')
g = g.map_dataframe(sns.barplot,"RY","Value",hue='Part_of_prog',ci=None,palette = sns.color_palette("bright"))
for ax in g.axes.ravel():
    ax.legend()
plt.show()

产生以下图形：现在，我想将条形图更改为堆积条形图。我想使用构面网格生成类似的布局。我使用以下代码生成图形，但出现错误：

  g = sns.FacetGrid(df,col='Direction')
    g = g.map_dataframe(df[['Value','RY']].plot.bar(stacked=True),hue='Part_of_prog',palette = sns.color_palette("bright"))
    for ax in g.axes.ravel():
        ax.legend()
    plt.show()

有人可以指导我如何在facegrid中生成堆积条形图吗？

Answer 1

正如在现在已删除的评论/答案中提到的那样，seaborn是一个自以为是的数据可视化库，这意味着它可以轻松地创建某些（主要是统计的）图形类型，而不是通用的“构建自己的工具”。根据其作者说，堆积的条形不属于scope。

因此，我建议您创建使用Bokeh之后的图形类型。数据将需要以一种稍微的不同方式进行结构化（必须将堆栈变量旋转到列中），并且还有一点要编写的代码，但这就是这样做的代价。更大的灵活性。

首先，进行数据操作：

df["Part_of_prog"] = df["Part_of_prog"].astype(str)
df["RY"] = df["RY"].astype(str)

wide_df = (df
.pivot_table(index=["Direction", "RY"], columns="Part_of_prog", values="Value", aggfunc=sum)
.reset_index()
.rename_axis(None, axis="columns")
)

现在是可视化部分：

from bokeh.io import output_notebook, output_file, show
from bokeh.plotting import figure
from bokeh.models import ColumnDataSource, CDSView, BooleanFilter, Range1d
from bokeh.layouts import row

output_notebook() # for Jupyter Notebooks

# shared ranges for the 3 plots
x_range = df["RY"].unique()
y_range = Range1d(0, 110)

colors = ["#c9d9d3", "#718dbf"]

# shared source for all plots to allow linked brushing/panning if you need it
source = ColumnDataSource(wide_df)

# filtered views for each of the facets
east_view = CDSView(source=source, filters=[BooleanFilter(wide_df["Direction"] == "East")])
north_view = CDSView(source=source, filters=[BooleanFilter(wide_df["Direction"] == "North")])
south_view = CDSView(source=source, filters=[BooleanFilter(wide_df["Direction"] == "South")])

# plots are broadly similar, but we'll use a facet object to modify things like titles
def generate_facet_plot(facet):
    
    p = figure(x_range=x_range, y_range=y_range, plot_height=250, plot_width=200,
               title=facet["title"], toolbar_location=None, tools="")

    p.vbar_stack(stackers=["0","1"], x="RY", color=colors, width=0.9,
                 source=source, view=facet["view"])
    
    p.yaxis.visible = facet["y_axis_visible"]
    p.xgrid.visible = False
    p.y_range.start = 0

    return p

# our configuration object for facets
facets = {
    "east": {
        "title": "East",
        "view" : east_view,
        "y_axis_visible": True
    },
    "north": {
        "title": "North",
        "view" : north_view,
        "y_axis_visible": False
    },
    "south": {
        "title": "South",
        "view" : south_view,
        "y_axis_visible": False
    },
}

plots = []

for facet in facets.keys():
    plots.append(generate_facet_plot(facets[facet]))
  
# row layout; other options are available
p = row(plots)
show(p)

结果：

Bokeh documentation中有许多样式（带有示例）。

Python：在Facet网格中绘制堆积的条形图

1 个答案: