找關係

Question

我在熊貓有一個數據幀，relation_between_countries區別：

country_from country_to points 
1 Albania  Austria  10 
2 Denmark  Austria  5 
3 Austria  Albania  2 
4 Greece   Norway  4 
5 Norway   Greece  5 

我試圖讓關係的點之間的差異，因爲這：

country_from_or_to country_to_or_from difference 
Albania    Austria    8 
Denmark    Austria    
Greece    Norway    -1 

你有什麼想法怎麼辦？

Answer 1

使用DataFrameGroupBy.diff：

cols = ['country_from','country_to'] 
#sort values in columns 
df[cols] = df[cols].apply(sorted, axis=1) 
#get difference 
df['difference'] = df.groupby(cols)['points'].diff(-1) 
print (df) 
    country_from country_to points difference 
1  Albania Austria  10   8.0 
2  Austria Denmark  5   NaN 
3  Albania Austria  2   NaN 
4  Greece  Norway  4  -1.0 
5  Greece  Norway  5   NaN 

也可以代替NaN空的空間，但在右列的混合值 - 數字處理字符串，因此一些函數可以返回奇怪的輸出：

cols = ['country_from','country_to'] 
df[cols] = df[cols].apply(sorted, axis=1) 
df['difference'] = df.groupby(cols)['points'].diff(-1).fillna('') 
print (df) 
    country_from country_to points difference 
1  Albania Austria  10   8 
2  Austria Denmark  5   
3  Albania Austria  2   
4  Greece  Norway  4   -1 
5  Greece  Norway  5