Question

我有一个数据框，里面有很多json数据。

例如：

{"serial": "000000001fb105ea", "sensorType": "acceleration", "data": [1603261123.328814, 0.171875, -0.9609375, 0.0234375]}
{"serial": "000000001fb105ea", "sensorType": "acceleration", "data": [1603261125.0605137, 0.0859375, -0.984375, 0.0]}
{"serial": "000000001fb105ea", "sensorType": "strain", "data": [1603261126.3532753, 0.9649793604217437]}
{"serial": "000000001fb105ea", "sensorType": "acceleration", "data": [1603261127.6988888, 0.0390625, -1.0, 0.125]}
{"serial": "000000001fb105ea", "sensorType": "acceleration", "data": [1603261128.8530502, 0.078125, -0.9921875, 0.0]}

数据有两种类型：应变传感器和加速度传感器。

我想解析这些json数据并转换为普通格式。我只需要json对象的数据部分。结果，我应该为Data中的每个值有4列。

Date: 21.20.2020:09:18:46    x:0.171875     y:-0.9609375    z:0.0234375

我尝试了json_normalize，但出现了此错误。

AttributeError: 'str' object has no attribute 'itervalues'

如何将数据部分解析为4列数据框？

谢谢。

Answer 1

如果输入数据位于json文件中，请使用：

cols = ['Date','x','y','z']
df = pd.DataFrame(pd.read_json('json.json', lines=True)['data'].tolist(), columns=cols)
df['Date'] = pd.to_datetime(df['Date'], unit='s')
print (df)
                           Date         x         y         z
0 2020-10-21 06:18:43.328814030  0.171875 -0.960938  0.023438
1 2020-10-21 06:18:45.060513735  0.085938 -0.984375  0.000000
2 2020-10-21 06:18:46.353275299  0.964979       NaN       NaN
3 2020-10-21 06:18:47.698888779  0.039062 -1.000000  0.125000
4 2020-10-21 06:18:48.853050232  0.078125 -0.992188  0.000000

如果输入为DataFrame，且列为col：

cols = ['Date','x','y','z']
df = pd.DataFrame(pd.json_normalize(df['col'])['data'].tolist(), columns=cols)
df['Date'] = pd.to_datetime(df['Date'], unit='s')
print (df)
                           Date         x         y         z
0 2020-10-21 06:18:43.328814030  0.171875 -0.960938  0.023438
1 2020-10-21 06:18:45.060513735  0.085938 -0.984375  0.000000
2 2020-10-21 06:18:46.353275299  0.964979       NaN       NaN
3 2020-10-21 06:18:47.698888779  0.039062 -1.000000  0.125000
4 2020-10-21 06:18:48.853050232  0.078125 -0.992188  0.000000

编辑：

像.xls这样个人保存csv并不是一个好主意，因为这样read_excel会引发奇怪的错误，但是您可以使用：

import ast

df = pd.read_csv('15-10-2020-OO.xls')

cols = ['Date','x','y','z']

data = [x['data'] for x in df['Data'].apply(ast.literal_eval)]
df = pd.DataFrame(data, columns=cols)
df['Date'] = pd.to_datetime(df['Date'], unit='s')
print (df)
                              Date         x         y         z
0    2020-10-15 07:21:16.159236193  0.085938 -0.972656  0.003906
1    2020-10-15 07:21:17.597931385  0.089844 -0.968750  0.003906
2    2020-10-15 07:21:18.838171959  0.089844 -0.972656  0.003906
3    2020-10-15 07:21:20.338105917  0.085938 -0.972656  0.003906
4    2020-10-15 07:21:21.768864155  0.089844 -0.984375  0.003906
                           ...       ...       ...       ...
8457 2020-10-15 08:59:57.907007933  0.085938 -0.972656  0.003906
8458 2020-10-15 08:59:58.371274233  0.089844 -0.976562  0.003906
8459 2020-10-15 08:59:58.833237648  0.085938 -0.976562  0.003906
8460 2020-10-15 08:59:59.313337088  1.517057       NaN       NaN
8461 2020-10-15 08:59:59.863240004  0.089844 -0.968750  0.007812

[8462 rows x 4 columns]

如何将json数据帧转换为普通数据帧？

1 个答案: