如何基于DataFrame绘制分组堆叠条形图并避免重叠

我用以下代码绘制堆叠条形图时，两组图出现重叠：

fig, ax = plt.subplots(figsize=(8,6))
df = pd.concat((data.assign(source=name) for data, name in zip([train_ss_freq_df, train_ss_freq_df], ['train', 'blind'])))
                
df[df['source']=='train'].plot(kind='bar', stacked=True, color=sns.color_palette("crest", 3),ax=ax)
df[df['source']=='blind'].plot(kind='bar', stacked=True, color=sns.color_palette("crest", 3),ax=ax)

使用的数据集如下：

,H,E,C
A,0.039065342828923426,0.014685241981597963,0.026069553464677445
R,0.023860269272755627,0.011930134636377814,0.017492332484275095
N,0.012605915683318605,0.007381608358891719,0.02365233664292769
D,0.018817902999428187,0.007589540988719654,0.03285335551281385
C,0.004028694702916255,0.002729115766491657,0.004470551541300619
Q,0.01886988615688517,0.007849456776004574,0.013437646202630348
E,0.032983313406456306,0.010578572542496232,0.020975204033893018
G,0.013827519883557727,0.013437646202630348,0.04777252170296824
H,0.008733170452773302,0.006004054686281644,0.011150387274523055
I,0.020793262982793576,0.021209128242449447,0.012969797785517493
L,0.04384779331496595,0.02666735977543276,0.024899932421895307
K,0.02352237874928523,0.010084732546654884,0.021832926131933255
M,0.007797473618547591,0.004678484171128554,0.005354265218069345
F,0.01465925040286947,0.013749545147372252,0.011150387274523055
P,0.009201018869886156,0.0047304673285855385,0.03246348183188647
S,0.017856214586473983,0.013229713572802412,0.028278837656599262
T,0.015387014607267246,0.01614077039039351,0.023990227166398086
W,0.006419919945937516,0.004600509434943078,0.004418568383843634
Y,0.01268389041950408,0.012943806206789,0.011202370431980038
V,0.02139106929354889,0.032151582887144564,0.016842543016062795

,H,E,C
A,0.04221830985915493,0.014446680080482898,0.02841549295774648
R,0.02193158953722334,0.011006036217303823,0.019577464788732395
N,0.010206237424547284,0.005719315895372234,0.023712273641851106
D,0.016267605633802817,0.007223340040241449,0.03400905432595573
C,0.0037927565392354124,0.0037374245472837023,0.005960764587525151
Q,0.017751509054325956,0.0064285714285714285,0.014305835010060362
E,0.03344567404426559,0.010326961770623743,0.026398390342052314
G,0.01176056338028169,0.011066398390342052,0.05057344064386318
H,0.00698692152917505,0.005060362173038229,0.01062374245472837
I,0.02187122736418511,0.02117706237424547,0.01375251509054326
L,0.043470824949698186,0.022062374245472836,0.028118712273641853
K,0.024170020120724348,0.011036217303822938,0.024622736418511065
M,0.009476861167002013,0.004592555331991952,0.007339034205231389
F,0.01391851106639839,0.011956740442655935,0.013083501006036218
P,0.007027162977867203,0.004964788732394366,0.033591549295774646
S,0.015774647887323943,0.01119718309859155,0.030653923541247484
T,0.015020120724346076,0.014507042253521127,0.024899396378269618
W,0.004964788732394366,0.003958752515090543,0.004678068410462776
Y,0.011156941649899397,0.010789738430583501,0.011091549295774649
V,0.023586519114688127,0.02992957746478873,0.018606639839034204

我已经将数据集合并并添加source字段区分组别，但两次调用plot会导致两组图重叠，不知道怎么设置参数让它们并排显示。

问题出在两次调用plot时，pandas默认会使用相同的x轴坐标绘制条形，所以两组图重叠。要实现并排，需要手动调整两组条形的x位置，具体步骤如下：

1. 拆分数据集，获取两组数据的索引（即氨基酸类别）
2. 设置条形宽度，计算两组条形的x坐标（一组左移，一组右移）
3. 分别绘制堆叠条形，手动指定x坐标
4. 调整x轴标签和图例

修改后的代码：

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

# 拆分并标记数据集（修正原代码中重复传入train_ss_freq_df的错误）
df_train = train_ss_freq_df.assign(source='train')
df_blind = blind_ss_freq_df.assign(source='blind')

categories = df_train.index  # 获取氨基酸类别
width = 0.35  # 设置单组条形宽度
x = np.arange(len(categories))  # 基础x坐标

fig, ax = plt.subplots(figsize=(12,6))  # 加宽画布避免标签重叠

# 绘制train组堆叠条形
train_data = df_train[['H','E','C']]
bottom = np.zeros(len(categories))
for i, col in enumerate(train_data.columns):
    ax.bar(x - width/2, train_data[col], width, bottom=bottom, 
           color=sns.color_palette("crest", 3)[i], label=f'train_{col}')
    bottom += train_data[col].values

# 绘制blind组堆叠条形
blind_data = df_blind[['H','E','C']]
bottom = np.zeros(len(categories))
for i, col in enumerate(blind_data.columns):
    ax.bar(x + width/2, blind_data[col], width, bottom=bottom, 
           color=sns.color_palette("crest", 3)[i], label=f'blind_{col}')
    bottom += blind_data[col].values

# 设置图表样式
ax.set_xticks(x)
ax.set_xticklabels(categories)
ax.set_ylabel('Frequency')
ax.set_title('Amino Acid Frequency Distribution (Train vs Blind)')
# 整理图例避免重复
handles, labels = ax.get_legend_handles_labels()
unique_legend = dict(zip(labels, handles))
ax.legend(unique_legend.values(), unique_legend.keys(), bbox_to_anchor=(1.05, 1), loc='upper left')

plt.tight_layout()
plt.show()

关键说明： #

• 用x - width/2和x + width/2分别设置两组条形的位置，实现并排效果
• 手动控制bottom参数实现堆叠（pandas的stacked=True内部逻辑一致）
• 加宽画布并调整图例位置，避免标签和图例重叠
• 修正了原代码中重复传入同一数据集的错误

内容的提问来源于stack exchange，提问作者Caterina