Seaborn regplot图例异常：图例显示圆点而非对应回归线

问题描述 #

使用Seaborn的regplot绘制自然灾害事件数据的回归线时，图表图例显示为圆点，无法对应各灾害类型的回归线；但改用lineplot时图例显示正常。代码如下：

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

df = pd.read_csv(r'C:\Users\ASUS\Downloads\number-of-natural-disaster-events.csv')
df.drop(['Code'], axis = 1)
df_pivot = df.pivot(index='Year',columns='Entity',values='Number of reported natural disasters (reported disasters)')
df_pivot = df_pivot.drop(['Impact'],axis=1)
df_pivot = df_pivot.fillna(0)
df_pivot = df_pivot.reset_index()

fig, ax = plt.subplots(1, 1, sharey=True, sharex=True, figsize=(20, 10))
legend = []
for col in df_pivot.columns:
    if col != "Year":
       sns.regplot(ax=ax,data=df_pivot,x="Year", y=col)
       legend.append(col)
       ax.legend(legend,bbox_to_anchor=(0.5, -0.1),fancybox=True,loc='upper center',shadow=True,
                   ncol=2)

问题原因 #

1. regplot默认同时绘制散点数据和拟合回归线，Matplotlib默认提取散点的标记作为图例项，而非回归线的线条，因此图例显示为圆点。
2. 循环中每次调用ax.legend()会重复生成图例，覆盖之前的设置，导致图例与线条无法正确对应。
3. 原代码中df.drop(['Code'], axis=1)未赋值给df，会导致后续处理仍包含Code列，属于潜在问题。

解决方法 #

方法一：仅绘制回归线（隐藏散点） #

如果不需要显示原始散点，只保留回归线，可在regplot中设置scatter=False，同时给每个回归线指定label参数，最后统一生成图例：

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

df = pd.read_csv(r'C:\Users\ASUS\Downloads\number-of-natural-disaster-events.csv')
# 修正drop操作的赋值问题
df = df.drop(['Code'], axis = 1)
df_pivot = df.pivot(index='Year',columns='Entity',values='Number of reported natural disasters (reported disasters)')
df_pivot = df_pivot.drop(['Impact'],axis=1)
df_pivot = df_pivot.fillna(0)
df_pivot = df_pivot.reset_index()

fig, ax = plt.subplots(1, 1, sharey=True, sharex=True, figsize=(20, 10))
for col in df_pivot.columns:
    if col != "Year":
       # 隐藏散点，指定label对应灾害类型
       sns.regplot(ax=ax, data=df_pivot, x="Year", y=col, scatter=False, label=col)
# 统一生成底部图例
ax.legend(bbox_to_anchor=(0.5, -0.1), fancybox=True, loc='upper center', shadow=True, ncol=2)
plt.show()

方法二：保留散点+回归线，生成对应图例 #

如果需要同时显示散点和回归线，需手动提取每个regplot返回的线条与散点元素，组合成匹配的图例项：

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

df = pd.read_csv(r'C:\Users\ASUS\Downloads\number-of-natural-disaster-events.csv')
df = df.drop(['Code'], axis = 1)
df_pivot = df.pivot(index='Year',columns='Entity',values='Number of reported natural disasters (reported disasters)')
df_pivot = df_pivot.drop(['Impact'],axis=1)
df_pivot = df_pivot.fillna(0)
df_pivot = df_pivot.reset_index()

fig, ax = plt.subplots(1, 1, sharey=True, sharex=True, figsize=(20, 10))
legend_elements = []
for col in df_pivot.columns:
    if col != "Year":
       # 获取regplot返回的散点和线条对象
       scatter, line = sns.regplot(ax=ax, data=df_pivot, x="Year", y=col)
       # 组合线条与散点为图例项
       legend_elements.append(plt.Line2D([], [], color=line.get_color(), marker=scatter.get_markers()[0], 
                                         linestyle='-', label=col))
# 用手动创建的图例元素生成对应图例
ax.legend(handles=legend_elements, bbox_to_anchor=(0.5, -0.1), fancybox=True, loc='upper center', shadow=True, ncol=2)
plt.show()

内容的提问来源于stack exchange，提问作者mobly elliot