13

Pareto 是 Excel 和 Tableu 中非常流行的图表。在 excel 中,我们可以轻松绘制帕累托图,但我发现在 Python 中绘制图表并不容易。

我有一个像这样的熊猫数据框:

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

df = pd.DataFrame({'country': [177.0, 7.0, 4.0, 2.0, 2.0, 1.0, 1.0, 1.0]})
df.index = ['USA', 'Canada', 'Russia', 'UK', 'Belgium', 'Mexico', 'Germany', 'Denmark']
print(df)

         country
USA        177.0
Canada       7.0
Russia       4.0
UK           2.0
Belgium      2.0
Mexico       1.0
Germany      1.0
Denmark      1.0

如何绘制帕累托图?可能使用 pandas、seaborn、matplotlib 等?

到目前为止,我已经能够制作降序条形图。但是仍然需要将累积总和线图放在它们之上。

我的尝试: df.sort_values(by='country',ascending=False).plot.bar()

所需地块:

4

5 回答 5

27

您可能希望创建一个包含百分比的新列,并将一列绘制为条形图,另一列绘制为双轴中的折线图。

import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.ticker import PercentFormatter

df = pd.DataFrame({'country': [177.0, 7.0, 4.0, 2.0, 2.0, 1.0, 1.0, 1.0]})
df.index = ['USA', 'Canada', 'Russia', 'UK', 'Belgium', 'Mexico', 'Germany', 'Denmark']
df = df.sort_values(by='country',ascending=False)
df["cumpercentage"] = df["country"].cumsum()/df["country"].sum()*100


fig, ax = plt.subplots()
ax.bar(df.index, df["country"], color="C0")
ax2 = ax.twinx()
ax2.plot(df.index, df["cumpercentage"], color="C1", marker="D", ms=7)
ax2.yaxis.set_major_formatter(PercentFormatter())

ax.tick_params(axis="y", colors="C0")
ax2.tick_params(axis="y", colors="C1")
plt.show()

在此处输入图像描述

于 2018-12-02T09:24:25.870 回答
5

另一种方法是使用secondary_y参数而不使用twinx()

df['pareto'] = 100 *df.country.cumsum() / df.country.sum()
fig, axes = plt.subplots()
ax1 = df.plot(use_index=True, y='country',  kind='bar', ax=axes)
ax2 = df.plot(use_index=True, y='pareto', marker='D', color="C1", kind='line', ax=axes, secondary_y=True)
ax2.set_ylim([0,110])

在此处输入图像描述

该参数use_index=True是必需的,因为在这种情况下您index是您的x轴。否则你可以使用x='x_Variable'.

于 2020-01-13T22:45:05.010 回答
5

pandas.dataframe 的帕累托图

import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.ticker import PercentFormatter


def _plot_pareto_by(df_, group_by, column):

    df = df_.groupby(group_by)[column].sum().reset_index()
    df = df.sort_values(by=column,ascending=False)

    df["cumpercentage"] = df[column].cumsum()/df[column].sum()*100


    fig, ax = plt.subplots(figsize=(20,5))
    ax.bar(df[group_by], df[column], color="C0")
    ax2 = ax.twinx()
    ax2.plot(df[group_by], df["cumpercentage"], color="C1", marker="D", ms=7)
    ax2.yaxis.set_major_formatter(PercentFormatter())

    ax.tick_params(axis="y", colors="C0")
    ax2.tick_params(axis="y", colors="C1")

    for tick in ax.get_xticklabels():
        tick.set_rotation(45)
    plt.show()

在此处输入图像描述

于 2020-05-20T12:01:55.257 回答
2

ImportanceOfBeingErnest 代码的更通用版本:

def create_pareto_chart(df, by_variable, quant_variable):
    df.index = by_variable
    df["cumpercentage"] = quant_variable.cumsum()/quant_variable.sum()*100

    fig, ax = plt.subplots()
    ax.bar(df.index, quant_variable, color="C0")
    ax2 = ax.twinx()
    ax2.plot(df.index, df["cumpercentage"], color="C1", marker="D", ms=7)
    ax2.yaxis.set_major_formatter(PercentFormatter())

    ax.tick_params(axis="y", colors="C0")
    ax2.tick_params(axis="y", colors="C1")
    plt.show()

并且这个也包括根据阈值分组的帕累托。例如:如果您将其设置为 70,它将把超过 70 的少数民族归为一组,称为“其他”。

def create_pareto_chart(by_variable, quant_variable, threshold):

    total=quant_variable.sum()
    df = pd.DataFrame({'by_var':by_variable, 'quant_var':quant_variable})
    df["cumpercentage"] = quant_variable.cumsum()/quant_variable.sum()*100
    df = df.sort_values(by='quant_var',ascending=False)
    df_above_threshold = df[df['cumpercentage'] < threshold]
    df=df_above_threshold
    df_below_threshold = df[df['cumpercentage'] >= threshold]
    sum = total - df['quant_var'].sum()
    restbarcumsum = 100 - df_above_threshold['cumpercentage'].max()
    rest = pd.Series(['OTHERS', sum, restbarcumsum],index=['by_var','quant_var', 'cumpercentage'])
    df = df.append(rest,ignore_index=True)
    df.index = df['by_var']
    df = df.sort_values(by='cumpercentage',ascending=True)

    fig, ax = plt.subplots()
    ax.bar(df.index, df["quant_var"], color="C0")
    ax2 = ax.twinx()
    ax2.plot(df.index, df["cumpercentage"], color="C1", marker="D", ms=7)
    ax2.yaxis.set_major_formatter(PercentFormatter())

    ax.tick_params(axis="x", colors="C0", labelrotation=70)
    ax.tick_params(axis="y", colors="C0")
    ax2.tick_params(axis="y", colors="C1")

    plt.show()
于 2018-12-13T12:17:20.490 回答
1

这是我使用 pandas 和 plotly 的 Pareto 图版本。您可以将任何集合与未分组的数据一起使用。让我们从这个例子的数据开始:

import numpy as np

data = np.random.choice(['USA', 'Canada', 'Russia', 'UK', 'Belgium',
                                'Mexico', 'Germany', 'Denmark'], size=500,
                                 p=[0.43, 0.14, 0.23, 0.07, 0.04, 0.01, 0.03, 0.05])

图表创建:

import pandas as pd
import plotly.graph_objects as go


def pareto_chart(collection):
    collection = pd.Series(collection)
    counts = (collection.value_counts().to_frame('counts')
              .join(collection.value_counts(normalize=True).cumsum().to_frame('ratio')))

    fig = go.Figure([go.Bar(x=counts.index, y=counts['counts'], yaxis='y1', name='count'),
                     go.Scatter(x=counts.index, y=counts['ratio'], yaxis='y2', name='cumulative ratio',
                                hovertemplate='%{y:.1%}', marker={'color': '#000000'})])

    fig.update_layout(template='plotly_white', showlegend=False, hovermode='x', bargap=.3,
                      title={'text': 'Pareto Chart', 'x': .5}, 
                      yaxis={'title': 'count'},
                      yaxis2={'rangemode': "tozero", 'overlaying': 'y',
                              'position': 1, 'side': 'right',
                              'title': 'ratio',
                              'tickvals': np.arange(0, 1.1, .2),
                              'tickmode': 'array',
                              'ticktext': [str(i) + '%' for i in range(0, 101, 20)]})

    fig.show()

结果: 在此处输入图像描述

于 2021-07-24T09:55:41.167 回答