python - 如何在循环中动态更新绘图？

Question

我有以下代码片段，我想以一种方式扩展它，使每个循环的数据都绘制在同一 Canvas 上，而不是将每个循环绘制到不同的 Canvas 上。

for level in range(len(result)):
  sizes = result[level].values()
  distribution=pd.DataFrame(Counter(sizes).items(), columns=['community size','number of communities'])
  distribution.plot(kind='scatter', x='community size', y='number of communities')

在最佳情况下，我还希望根据原始数据对散点图中的点进行颜色编码(属于一个循环的数据的点以相同颜色着色)。

我对 matplotlib 和 pandas 都比较陌生，非常感谢 andy 的帮助。

Answer 1

无需多次调用 plot，您可以将整个数据集构建为一个 DataFrame 然后你只需要调用一次 plot。

开始于

result = [{0: 21, 1: 7, 2: 67, 3: 12, 4: 15, 5: 7, 6: 54, 7: 49, 8: 50, 9: 31,
           10: 6, 11: 2, 12: 8, 13: 2, 14: 2, 15: 1, 16: 35, 17: 2, 18: 1, 19:
           4, 20: 2, 21: 4, 22: 3, 23: 1, 24: 1, 25: 1, 26: 1, 27: 1, 28: 1, 
           29: 1}, 
          {0: 2, 1: 5, 2: 2, 3: 3, 4: 1, 5: 2, 6: 3, 7: 2, 8: 1, 9: 1, 10: 1,
           11: 1, 12: 1, 13: 1, 14: 1, 15: 1, 16: 1, 17: 1}]

您可以构建一个包含列 level 和 size 的 DataFrame:

df = pd.DataFrame([(level,val) for level, dct in enumerate(result) 
                   for val in dct.values()],
                  columns=['level', 'size'])

看起来像这样:

    level  size
0       0    21
1       0     7
2       0    67
...
45      1     1
46      1     1
47      1     1

现在我们可以按级别分组，并计算每个组中每个大小的项目有多少:

size_count = df.groupby(['level'])['size'].apply(lambda x: x.value_counts())
# level    
# 0      1      9
#        2      5
#        7      2
# ...
# 1      1     11
#        2      4
#        3      2
#        5      1
# dtype: int64

groupby/apply以上返回一个 pd.Series。为了使它成为一个 DataFrame，我们可以通过调用 reset_index() 将索引级别的值放入列中，然后为列分配列名:

size_count = size_count.reset_index()
size_count.columns = ['level', 'community size', 'number of communities']

现在可以生成所需的绘图

size_count.plot(kind='scatter', x='community size', y='number of communities', 
                s=100, c='level')

s=100 控制点的大小，c='level' 告诉 plot 根据中的值给点着色级别列。

---

import pandas as pd
import matplotlib.pyplot as plt

result = [{0: 21, 1: 7, 2: 67, 3: 12, 4: 15, 5: 7, 6: 54, 7: 49, 8: 50, 9: 31,
           10: 6, 11: 2, 12: 8, 13: 2, 14: 2, 15: 1, 16: 35, 17: 2, 18: 1, 19:
           4, 20: 2, 21: 4, 22: 3, 23: 1, 24: 1, 25: 1, 26: 1, 27: 1, 28: 1, 
           29: 1}, 
          {0: 2, 1: 5, 2: 2, 3: 3, 4: 1, 5: 2, 6: 3, 7: 2, 8: 1, 9: 1, 10: 1,
           11: 1, 12: 1, 13: 1, 14: 1, 15: 1, 16: 1, 17: 1}]

df = pd.DataFrame([(level,val) for level, dct in enumerate(result) 
                   for val in dct.values()],
                  columns=['level', 'size'])
size_count = df.groupby(['level'])['size'].apply(lambda x: x.value_counts())
size_count = size_count.reset_index()
size_count.columns = ['level', 'community size', 'number of communities']
cmap = plt.get_cmap('jet')
size_count.plot(kind='scatter', x='community size', y='number of communities', 
                s=100, c='level', cmap=cmap)
plt.show()

如果有几十个级别，使用颜色条可能是合适的。

---

另一方面，如果只有几个级别，使用图例会使 更有意义。在那种情况下，为每个调用一次 plot 会更方便 level 值，因为制作图例的 matplotlib 代码设置为 每个地 block 一个图例条目:

import pandas as pd
import matplotlib.pyplot as plt

result = [{0: 21, 1: 7, 2: 67, 3: 12, 4: 15, 5: 7, 6: 54, 7: 49, 8: 50, 9: 31,
           10: 6, 11: 2, 12: 8, 13: 2, 14: 2, 15: 1, 16: 35, 17: 2, 18: 1, 19:
           4, 20: 2, 21: 4, 22: 3, 23: 1, 24: 1, 25: 1, 26: 1, 27: 1, 28: 1, 
           29: 1}, 
          {0: 2, 1: 5, 2: 2, 3: 3, 4: 1, 5: 2, 6: 3, 7: 2, 8: 1, 9: 1, 10: 1,
           11: 1, 12: 1, 13: 1, 14: 1, 15: 1, 16: 1, 17: 1}]

df = pd.DataFrame([(level,val) for level, dct in enumerate(result) 
                   for val in dct.values()],
                  columns=['level', 'size'])
groups = df.groupby(['level'])
fig, ax = plt.subplots()
for level, grp in groups:
    size_count = grp['size'].value_counts()
    ax.plot(size_count.index, size_count, markersize=12, marker='o', 
            linestyle='', label='level {}'.format(level))
ax.legend(loc='best', numpoints=1)
ax.set_xlabel('community size')
ax.set_ylabel('number of communities')
ax.grid(True)
plt.show()