重塑堆叠 Pandas DataFrame

Question

我有以下 DataFrame df1:

 df1 = pd.DataFrame(np.random.rand(4,2), columns = {"var1", "var2"})
 df1["inst"] = ["A", "A", "B", "B"]
 df1.set_index("inst", inplace = True)
 df1 = df1.stack()

ipdb> df1
      inst
      A     var1    0.191094
            var2    0.100821
            var1    0.251331
            var2    0.528787
      B     var1    0.806549
            var2    0.638217
            var1    0.233541
            var2    0.905737

我想重塑 df1 这样

ipdb> df1
              A           B
      var1    0.191094    0.806549
      var2    0.100821    0.638217
      var1    0.251331    0.233541
      var2    0.528787    0.905737

我尝试获取 df1 的值并使用 reshape 函数重塑它们，但没有成功：

ipdb> df1.values
      array([ 0.19109431,  0.10082081,  0.25133097,  0.52878702,  
              0.80654863,  0.63821703,  0.23354052,  0.90573699])

ipdb> df1.values.reshape(4,2)
      array([[ 0.19109431,  0.10082081],
             [ 0.25133097,  0.52878702],
             [ 0.80654863,  0.63821703],
             [ 0.23354052,  0.90573699]])

Answer 1

使用cumcount + set_index and reshape by unstack and stack:

g = df1.groupby('inst').cumcount()
df1 = df1.set_index(["inst",g]).unstack(0).stack(0).reset_index(level=0, drop=True)
print (df1)
inst         A         B
var1  0.932293  0.214795
var2  0.503961  0.904046
var1  0.943864  0.232308
var2  0.398277  0.379333

Answer 2

试试这个：

values = np.transpose(df1.values.reshape(2, 4))
df2 = pd.DataFrame(data=values, index=['var1', 'var2', 'var1', 'var2'], columns=['A', 'B'])
print(df2)

重塑堆叠 Pandas DataFrame

Reshape stacked Pandas DataFrame

reshape

python-3.x

pandas