Python 如何防止pandas在JSON转换期间更改多索引数据帧结构？

我最初有以下数据帧：

cat   cat    1  1  1  1  1
      dog    0  0  0  0  0
      fox    0  0  0  0  0
      jumps  1  1  1  1  1
      mouse  1  1  1  1  1
      over   1  1  1  1  1
      the    1  1  1  1  1
dog   cat    0  0  0  0  0
      dog    1  1  1  1  1
      fox    1  1  1  1  1
      jumps  1  1  1  1  1
      mouse  0  0  0  0  0
      over   1  1  1  1  1
      the    1  1  1  1  1
fox   cat    0  0  0  0  0
      dog    1  1  1  1  1
      fox    1  1  1  1  1
      jumps  1  1  1  1  1
      mouse  0  0  0  0  0
      over   1  1  1  1  1
      the    1  1  1  1  1
jumps cat    1  1  1  1  1
      dog    1  1  1  1  1
      fox    1  1  1  1  1
      jumps  1  1  1  1  1
      mouse  1  1  1  1  1
      over   1  0  1  1  1
      the    1  0  1  1  1
mouse cat    1  1  1  1  1
      dog    0  0  0  0  0
      fox    0  0  0  0  0
      jumps  1  1  1  1  1
      mouse  1  1  1  1  1
      over   1  1  1  1  1
      the    1  1  1  1  1
over  cat    1  1  1  1  1
      dog    1  1  1  1  1
      fox    1  1  1  1  1
      jumps  1  1  0  1  0
      mouse  1  1  1  1  1
      over   1  1  1  1  1
      the    1  0  1  1  1
the   cat    1  1  1  1  1
      dog    1  1  1  1  1
      fox    1  1  1  1  1
      jumps  1  1  0  1  1
      mouse  1  1  1  1  1
      over   1  1  0  1  0
      the    1  1  1  1  1

然后我使用以下方法将其转换为JSON格式：

JSON = df.reset_index().to_json()

new_df = pd.read_json(JSON)

重置索引，因为无法直接将多索引df转换为JSON

然后，我使用以下方法将其恢复为数据帧格式：

JSON = df.reset_index().to_json()

new_df = pd.read_json(JSON)

但现在当我打印它时，它显示以下数据帧：

0   1  1  1  1  1     cat     cat
1   0  0  0  0  0     cat     dog
10  1  1  1  1  1     dog   jumps
11  0  0  0  0  0     dog   mouse
12  1  1  1  1  1     dog    over
13  1  1  1  1  1     dog     the
14  0  0  0  0  0     fox     cat
15  1  1  1  1  1     fox     dog
16  1  1  1  1  1     fox     fox
17  1  1  1  1  1     fox   jumps
18  0  0  0  0  0     fox   mouse
19  1  1  1  1  1     fox    over
2   0  0  0  0  0     cat     fox
20  1  1  1  1  1     fox     the
21  1  1  1  1  1   jumps     cat
22  1  1  1  1  1   jumps     dog
23  1  1  1  1  1   jumps     fox
24  1  1  1  1  1   jumps   jumps
25  1  1  1  1  1   jumps   mouse
26  1  0  1  1  1   jumps    over
27  1  0  1  1  1   jumps     the
28  1  1  1  1  1   mouse     cat
29  0  0  0  0  0   mouse     dog
3   1  1  1  1  1     cat   jumps
30  0  0  0  0  0   mouse     fox
31  1  1  1  1  1   mouse   jumps
32  1  1  1  1  1   mouse   mouse
33  1  1  1  1  1   mouse    over
34  1  1  1  1  1   mouse     the
35  1  1  1  1  1    over     cat
36  1  1  1  1  1    over     dog
37  1  1  1  1  1    over     fox
38  1  1  0  1  0    over   jumps
39  1  1  1  1  1    over   mouse
4   1  1  1  1  1     cat   mouse
40  1  1  1  1  1    over    over
41  1  0  1  1  1    over     the
42  1  1  1  1  1     the     cat
43  1  1  1  1  1     the     dog
44  1  1  1  1  1     the     fox
45  1  1  0  1  1     the   jumps
46  1  1  1  1  1     the   mouse
47  1  1  0  1  0     the    over
48  1  1  1  1  1     the     the
5   1  1  1  1  1     cat    over
6   1  1  1  1  1     cat     the
7   0  0  0  0  0     dog     cat
8   1  1  1  1  1     dog     dog
9   1  1  1  1  1     dog     fox

---

正如你所看到的，结构已经完全改变了。在与JSON的转换过程中，有什么方法可以保持数据帧的结构完全相同吗？

这里的索引只是添加了sort\u索引

---

这不起作用，但df仍然在结构上保持不变damaged@Elisha512在您将之前的_索引重置为_json后，df仍然会在那里受到结构损坏，以便在不损坏结构的情况下将其保存为json？@Elisha512您知道您可以将其转换回右侧吗？@Elisha512 pd.read_json.sort_index.set_index['level_0'，'level_1']