将嵌套的字典列表展开为Pandas数据框架

Question

我正在阅读以下JSON结构：

{"response":
    {"GDUEACWF":
        {"2018-06-01":
            [{"groupwide_market":"Developed Markets",
            "weights":0.8794132316432903},
            {"groupwide_market":"Developed Markets",
            "weights":0.8794132316432903}],
        "2018-06-02":
            [{"groupwide_market":"Developed Markets",  
            "weights":0.8794132316432903},
            {"groupwide_market":"Developed Markets",
            "weights":0.8794132316432903}]}}}

并且试图将其转换为以下格式的Pandas数据框。

|data_date  |groupwide_market  |weights
|2018-06-01 |Developed Markets |0.08794132316432903

我尝试通过使用以下代码循环遍历每个k,v对中的每个列表来实现此操作。虽然确实可以工作，但速度非常慢。100k行数据需要超过30分钟来生成。

df = pd.DataFrame()
#concatenating each line of the list within each dict cell
for k1,v1 in data['response'][mnemonic].items():
    for ele in v1:
        df_temp = pd.concat({k2: pd.Series(v2) for k2, v2 in ele.items()}).transpose()
        df_temp['data_date'] = k1
        df = df.append(df_temp,ignore_index=True)
df.columns = [x[0] for x in df.columns]

请问是否有更高效的方法来做这件事？我已经阅读了json_normalize的文档和示例，但无法在此情况下找出如何应用它。谢谢！

- min

Answer 1

假设有一个名为data的字典，我们可以按照如下方式进行操作：

import pandas as pd
pd.DataFrame([(date, *nodes.values()) for info in data["response"].values()
              for date, values in info.items() for nodes in values],
              columns=["date", "market", "weight"])

使用给定的响应作为输入，输出如下：

- Luca Cappelletti