将字典或元组列表转换为 DataFrame。操作 - 列，值 - 行

Question

我正在对 Facebook 进行 API 调用，其中一个字段是“操作”，它创建了一个字典，我想将其分解为单独的 DataFrame 列。我已经看到一些类似的问题，使用 pd.Series() 将它们映射到单独的列或 json.normalize() 中，但这些问题并不完全符合我的要求。

这是我将其放入数据框列之前的导出：

[<AdsInsights> {
    "actions": [
        {
            "action_type": "landing_page_view",
            "value": "292"
        },
        {
            "action_type": "comment",
            "value": "13"
        },
        {
            "action_type": "onsite_conversion.post_save",
            "value": "6"
        },
        {
            "action_type": "link_click",
            "value": "874"
        },
        {
            "action_type": "post",
            "value": "1"
        },
        {
            "action_type": "post_reaction",
            "value": "393"
        },
        {
            "action_type": "post_engagement",
            "value": "96"
        },
        {
            "action_type": "page_engagement",
            "value": "96"
        },
        {
            "action_type": "omni_activate_app",
            "value": "5"
        },
        {
            "action_type": "omni_app_install",
            "value": "2"
        },
        {
            "action_type": "omni_add_to_cart",
            "value": "75"
        },
        {
            "action_type": "add_to_wishlist",
            "value": "14"
        },
        {
            "action_type": "omni_purchase",
            "value": "4"
        },
        {
            "action_type": "omni_search",
            "value": "12"
        },
        {
            "action_type": "omni_view_content",
            "value": "15"
        }
    ]

然后我将其放入 DF 中，但该列变成了操作，它将所有这些数据放入一行，然后为每个项目重复它。我无法将 action_type 分解为列标题并将值分解为行。创建 DF 时，我得到了多行这些数据集。

当我制作数据框时，它看起来像这样：

    df[ad] df[actions]
0   ad1     [{'action_type': 'landing_page_view', 'value':...
1   ad2     [{'action_type': 'landing_page_view', 'value':...
2   ad3     [{'action_type': 'landing_page_view', 'value':...

我希望得到：

    df[ad] df[landing_page_view] df[comment] ...etc
0   ad1     292                      13
1   ad2     100                      8 
2   ad3     80                       9 

我试图挑选出我想要制作专栏的具体内容，但这不起作用：

df = all of the raw data from the API call
df = pd.DataFrame(df)
actions = df['actions']
def setcolumn(dict, key): 

    if dict.has_key(key): 
        df['key'] = 'value' 
    else: 
        print ("Not present")

setcolumn(actions, 'landing_page_view')

但这表示系列对象没有属性has_key。

感谢任何方向！

Answer 1

我稍微编辑了你的 API 结果，因为我不确定开头的标签是干什么用的。

解决方案是一个名为 from_records() 的简单 pandas 方法，它将列表转换为数据帧。以下是一些参考链接：

• https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.from_records.html
• https://pbpython.com/pandas-list-dict.html
• How to create pandas DataFrame with index from the list of tuples

import pandas as pd

# facebook API result as dictionary ... I removed the '<AdsInsights> ' at the beginning
list_facebook_api = {
    "actions": [
        {
            "action_type": "landing_page_view",
            "value": "292"
        },
        {
            "action_type": "comment",
            "value": "13"
        },
        {
            "action_type": "onsite_conversion.post_save",
            "value": "6"
        },
        {
            "action_type": "link_click",
            "value": "874"
        },
        {
            "action_type": "post",
            "value": "1"
        },
        {
            "action_type": "post_reaction",
            "value": "393"
        },
        {
            "action_type": "post_engagement",
            "value": "96"
        },
        {
            "action_type": "page_engagement",
            "value": "96"
        },
        {
            "action_type": "omni_activate_app",
            "value": "5"
        },
        {
            "action_type": "omni_app_install",
            "value": "2"
        },
        {
            "action_type": "omni_add_to_cart",
            "value": "75"
        },
        {
            "action_type": "add_to_wishlist",
            "value": "14"
        },
        {
            "action_type": "omni_purchase",
            "value": "4"
        },
        {
            "action_type": "omni_search",
            "value": "12"
        },
        {
            "action_type": "omni_view_content",
            "value": "15"
        }
    ]
}

# create a dataframe from the list inside the dictionary
# https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.from_records.html
# https://pbpython.com/pandas-list-dict.html
# https://stackoverflow.com/questions/44563707/how-to-create-pandas-dataframe-with-index-from-the-list-of-tuples
df = pd.DataFrame.from_records(list_facebook_api['actions'])

print(df.shape)

# print(df)
df