python - pandas 按条件频率对每个用户进行分组

Question

我有一个这样的数据框

df = pd.DataFrame({
    'User':['101','101','102','102','102'],
    'Product':['x','x','x','z','z'],
    'Country':['India,Brazil','India','India,Brazil,Japan','India,Brazil','Brazil']
})

我想通过用户明智地获得国家和产品组合计数，如下所示

首先拆分国家，然后结合产品并进行计数。

想要的输出：

Answer 1

这是在 SO 上结合其他答案的一种方法（这只是显示了搜索的力量：D）

import pandas as pd

df = pd.DataFrame({
    'User':['101','101','102','102','102'],
    'Product':['x','x','x','z','z'],
    'Country':['India,Brazil','India','India,Brazil,Japan','India,Brazil','Brazil']
})

# Making use of: https://stackoverflow.com/a/37592047/7386332
j = (df.Country.str.split(',', expand=True).stack()
                                           .reset_index(drop=True, level=1)
                                           .rename('Country'))
df = df.drop('Country', axis=1).join(j)

# Reformat to get desired Country_Product
df = (df.drop(['Country','Product'], 1)
      .assign(Country_Product=['_'.join(i) for i in zip(df['Country'], df['Product'])]))

df2 = df.groupby(['User','Country_Product'])['User'].count().rename('Count').reset_index()

print(df2)

回报：

  User Country_Product  count
0  101        Brazil_x      1
1  101         India_x      2
2  102        Brazil_x      1
3  102        Brazil_z      2
4  102         India_x      1
5  102         India_z      1
6  102         Japan_x      1

Answer 2

怎么样get_dummies

df.set_index(['User','Product']).Country.str.get_dummies(sep=',').replace(0,np.nan).stack().sum(level=[0,1,2])
Out[658]: 
User  Product        
101   x        Brazil    1.0
               India     2.0
102   x        Brazil    1.0
               India     1.0
               Japan     1.0
      z        Brazil    2.0
               India     1.0
dtype: float64