Pandas Dataframe - 用于分隔列的字典列表

Question

I have a dataframe, df, which has a column which is a list of dictionaries:

index action
0     [{'action_type': 'landing_page_view', 'value': '1'}, {'action_type': 'link_click', 'value': '1'}{'action_type': 'page_engagement', 'value': '1'}, {'action_type': 'post_engagement', 'value': '1'}]
1     [{'action_type': 'landing_page_view', 'value': '1'}, {'action_type': 'link_click', 'value': '1'}, {'action_type': 'page_engagement', 'value': '1'}, {'action_type': 'post_engagement', 'value': '1'}]
2     [{'action_type': 'video_view', 'value': '23'}, {'action_type': 'page_engagement', 'value': '23'}, {'action_type': 'post_engagement', 'value': '23'}]

I want to be able to extract the value from each dictionary and attribute it to its own column eg

index action_landing_page_view action_link_click action_page_engagement action_post_engagement action_video_view
0     1                        1                 1                      1                      0
1     1                        1                 1                      1                      0
2     0                        0                 23                     23                     23

I have tried df.apply(pd.Series) which splits out the dicts into seperate columns but with no column headers.

The dictionaries that are in my original dataframe do not follow the same order. eg the first dict for in row 1 starts with action_type "landing_page_view" whereas row 3 starts with "video_view".

Is it possible to attribute values to different columns based on the action_type in the dictionary?

Answer 1

You should first extract key-values for column name and column values, then constructing a dataframe based on them. (I think from the very first place it does not make sense to have such dictionaries. It makes sense to have them like this: {'landing_page_view': 1} )

index_action_extracted = [{} for i in range(len(index_action))]

for i in range(len(index_action)):
    list_ = index_action[i]
    for dict_ in list_:
        column_name = dict_['action_type']
        column_value = dict_['value']
        index_action_extracted[i].update({column_name: column_value})

df = pd.DataFrame(index_action_extracted).fillna(0)

In index_action_extracted we are removing redundant keys 'action_type' and 'value' and build a dictionary without those keys.

Answer 2

We can solve this using list comprehensions:

pd.DataFrame(
    [{e['action_type']: e['value']
      for e in l}
     for l in list(df['action'])]
).fillna(0)

Notice that I reconstructed your dataframe as follow:

import pandas as pd

df = pd.DataFrame([[
    [
        {'action_type': 'landing_page_view', 'value': '1'},
        {'action_type': 'link_click', 'value': '1'},
        {'action_type': 'page_engagement', 'value': '1'},
        {'action_type': 'post_engagement', 'value': '1'},
    ],
    [
        {'action_type': 'landing_page_view', 'value': '1'},
        {'action_type': 'link_click', 'value': '1'},
        {'action_type': 'page_engagement', 'value': '1'},
        {'action_type': 'post_engagement', 'value': '1'},
    ],
    [
        {'action_type': 'video_view', 'value': '23'},
        {'action_type': 'page_engagement', 'value': '23'},
        {'action_type': 'post_engagement', 'value': '23'},
    ]]]
)
df = df.transpose()
df = df[0].rename('action')
df = pd.DataFrame(df)

Answer 3

I return a list of dictionary items and then I map the dictionary items into the dataframe then transpose the dataframe

 df = pd.DataFrame([
    [
    {'action_type': 'landing_page_view', 'value': '1'},
    {'action_type': 'link_click', 'value': '1'},
    {'action_type': 'page_engagement', 'value': '1'},
    {'action_type': 'post_engagement', 'value': '1'},
],
[
    {'action_type': 'landing_page_view', 'value': '1'},
    {'action_type': 'link_click', 'value': '1'},
    {'action_type': 'page_engagement', 'value': '1'},
    {'action_type': 'post_engagement', 'value': '1'},
],
[
    {'action_type': 'video_view', 'value': '23'},
    {'action_type': 'page_engagement', 'value': '23'},
    {'action_type': 'post_engagement', 'value': '23'},
]
])

df2=pd.DataFrame(columns=['action_type','value'])
for key,dictList in df.iterrows():
    for key, dictAction in dictList.items():  #returns tuple (index, dict)
    #print(dictAction)
        if (dictAction is None)==False:
        
            keys=dictAction.keys()
            index=len(df2)
            for key in keys:
               value=dictAction[key]
               df2.loc[index,key]=value

df2=df2.T
print(df2.head())