![](/img/trans.png)
[英]How can I create a column in a dataframe using conditional logic on multiple columns in another dataframe python pandas?
[英]How can i create conditional column referring present columns of a dataframe and dictionary without using loop in python?
我有一個數據農場
import pandas as pd
df = pd.DataFrame({"type": ["A" ,"A1" ,"A" ,"A1","B" ],
"group": ["g1", "g2","g2","g2","g1"]})
我有一本字典
dic ={"AlphaA": {"A": {"g1":"A_GRP1", "g2":"A_GRP2"},
"A1": {"g1":"A1_GRP1", "g2":"A1_GRP2"}},
"AlphaB": {"B": {"g1":"B_GRP1", "g2":"B_GRP2"}},
}
我必須創建一個列名“值”,它將使用數據框和字典並獲取分配給它的值
申請條件:
第一行示例:
類型是“A”,因此引用字典鍵“AlphaA”
組是“g1
因此:
dictt["AlphaA"]["A"]["g1"] #would be the answer
所需輸出
final_df = pd.DataFrame({"type" : ["A" ,"A1" ,"A" ,"A1","B" ],
"group": ["g1", "g2","g2","g2","g1"],
"value": ["A_GRP1", "A1_GRP2", "A_GRP2",
"A1_GRP2", "B_GRP1"]})
我能夠使用循環來實現這一點,但它需要很多時間,
因此尋找一些快速的技術。
假設dic
輸入字典,您可以將字典值合並到單個字典中(在ChainMap
的幫助下),轉換為 DataFrame 並取消unstack
到 Series 並merge
:
from collections import ChainMap
s = pd.DataFrame(dict(ChainMap(*dic.values()))).unstack()
# without ChainMap
# d = {k: v for d in dic.values() for k,v in d.items()}
# pd.DataFrame(d).unstack()
out = df.merge(s.rename('value'), left_on=['type', 'group'], right_index=True)
輸出:
type group value
0 A g1 A_GRP1
1 A1 g2 A1_GRP2
3 A1 g2 A1_GRP2
2 A g2 A_GRP2
4 B g1 B_GRP1
將DataFrame.join
與通過字典理解從字典創建的 Series 一起使用:
d1 = {(k1, k2): v2 for k, v in d.items() for k1, v1 in v.items() for k2, v2 in v1.items()}
df = df.join(pd.Series(d1).rename('value'), on=['type','group'])
print (df)
type group value
0 A g1 A_GRP1
1 A1 g2 A1_GRP2
2 A g2 A_GRP2
3 A1 g2 A1_GRP2
4 B g1 B_GRP1
您可以刪除原始字典的外鍵並嘗試應用於行
d = {k:v for vs in d.values() for k, v in vs.items()}
df['value'] = (df.assign(value=df['type'].map(d))
.apply(lambda row: row['value'][row['group']], axis=1)
)
print(d)
{'A': {'g1': 'A_GRP1', 'g2': 'A_GRP2'}, 'A1': {'g1': 'A1_GRP1', 'g2': 'A1_GRP2'}, 'B': {'g1': 'B_GRP1', 'g2': 'B_GRP2'}}
print(df)
type group value
0 A g1 A_GRP1
1 A1 g2 A1_GRP2
2 A g2 A_GRP2
3 A1 g2 A1_GRP2
4 B g1 B_GRP1
聲明:本站的技術帖子網頁,遵循CC BY-SA 4.0協議,如果您需要轉載,請注明本站網址或者原文地址。任何問題請咨詢:yoyou2525@163.com.