How can I swap values within classes please?

As shown in this table:

- - - - - - - - - - Before - - - - - - - - - - - - - - - - - - After - - - - - - - - - -

I want to do this because it is over sampled data. It is very repetitive and this causes machine learning tools to over fit.

OK, try this out:

# 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

It's a bit grubby, and I'm sure there's a cleaner way to build that substitution map in perhaps a one-line list comprehension - but that should do the trick.

Alternately, there's the np.permute function which might bear some fruit in terms of adding some noise (albeit not by performing discrete swaps).

[edit] For testing, try a dataset with a bit less rigidity, here's an example of a more randomly generated one. Just edit out the columns you want to replace with fixed values if you want to impose some order in the dataset.

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")