pandas - 在类中交换值

Question

请问如何在类中交换值？

如下表所示：

- - - - - - - - - 之前 - - - - - - - - - - - - - - - - - 之后 - - - - - - - - -

我想这样做是因为它是过度采样的数据。这是非常重复的，这会导致机器学习工具过度拟合。

标签： pandasdiscrete-mathematics

Answer 1

好的，试试这个：

# Setup example dataframe
df = pd.DataFrame({"Class" : [1,2,1,3,1,2,1,3,1,2,1,3,1,2,1,3], 
                1:[1,1,1,0,1,0,1,0,1,0,1,0,1,0,1,1], 
                2:[0,1,0,0,0,1,0,0,0,1,0,0,0,1,0,0],
                3:[0,0,1,1,1,0,1,1,0,0,1,1,1,0,1,1],
                4:[1,0,1,1,1,0,1,1,1,0,1,1,1,0,1,1], 
                5:[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1],
                6:[0,0,0,1,0,0,0,1,0,0,0,1,0,0,0,1]}).set_index("Class")
# Do a filter on class, and store the positions/index of matching contents
class_to_edit=3
swappable_indices = np.where(df.index==class_to_edit)[0]
# Extract the column to edit
column_to_edit=1
column_values = df[column_to_edit].values

# Decide how many values to swap, and randomly assign swaps
# No guarantee here that the swaps will not contain the same values i.e. you could
# end up swapping 1's for 1's and 0's for 0's here - it's entirely random. 
number_of_swaps = 2
swap_pairs = np.random.choice(swappable_indices,number_of_swaps*2, replace=False)

# Using the swap pairs, build a map of substitutions, 
# starting with a vanilla no-swap map, then updating it with the generated swaps
swap_map={e:e for e in range(0,len(column_values))}
swap_map.update({swappable_indices[e]:swappable_indices[e+1] for e in range(0,len(swap_pairs),2)})
swap_map.update({swappable_indices[e+1]:swappable_indices[e] for e in range(0,len(swap_pairs),2)})

# Having built the swap-map, apply it to the data in the column, 
column_values=[column_values[swap_map[e]] for e,v in enumerate(column_values)]
# and then plug the column back into the dataframe
df[column_to_edit]=column_values

这有点肮脏，我确信有一种更简洁的方法可以在单行列表理解中构建替换映射 - 但这应该可以解决问题。

或者，还有一个np.permute函数，它可能会在增加一些噪音方面取得一些成果（尽管不是通过执行离散交换）。

[编辑] 为了进行测试，请尝试使用刚性较小的数据集，这是一个更随机生成的数据集的示例。如果您想在数据集中强加一些顺序，只需编辑出您想用固定值替换的列。

df = pd.DataFrame({"Class" : [1,2,1,3,1,2,1,3,1,2,1,3,1,2,1,3], 
            1:np.random.choice([0,1],16), 
            2:np.random.choice([0,1],16),
            3:np.random.choice([0,1],16),
            4:np.random.choice([0,1],16), 
            5:np.random.choice([0,1],16),
            6:np.random.choice([0,1],16)}).set_index("Class")