A sample input data frame
import pandas as pd
df_input = pd.DataFrame([[1.7, 0.2], [0.4, 0.93], [0.05, 0.96], [0.97, 0.68]], columns=["A", "B"])
This example has two columns whereas the real dataframe has 10. I want to sort each row in ascending order and then assign -1 to first 5 columns and +1 to next 5 columns. Sample output is as follows:-
df_output=pd.DataFrame([[1, -1], [-1, 1], [-1, 1], [1, -1]], columns=["A", "B"])
Please suggest the way forward.
You want np.argsort:
np.argsort(df_input, axis=1).replace(0, -1)
A B
0 1 -1
1 -1 1
2 -1 1
3 1 -1
To generalise to N rows:
v = np.where(np.argsort(df_input) >= df.shape[1] // 2, 1, -1)
df_output = pd.DataFrame(v)
print(df)
0 1 2 3 4 5 6 7 8 9
0 49 80 80 27 15 13 52 50 48 69
1 51 24 55 73 81 55 32 67 19 14
2 67 2 29 19 14 89 54 83 22 64
3 24 55 87 94 22 61 74 26 37 8
v = np.where(np.argsort(df_input) >= df.shape[1] // 2, 1, -1)
df_output = pd.DataFrame(v)
print(df_output)
0 1 2 3 4 5 6 7 8 9
0 1 -1 -1 1 -1 1 1 1 -1 -1
1 1 1 -1 1 -1 -1 1 1 -1 -1
2 -1 -1 -1 1 -1 1 1 -1 1 1
3 1 -1 -1 1 1 -1 1 1 -1 -1
Use numpy.where with np.argsort:
np.random.seed(111)
df_input = pd.DataFrame(np.random.randint(10, size=(10, 10)), columns=list('abcdefghij'))
print (df_input)
a b c d e f g h i j
0 6 8 3 6 6 7 1 8 3 4
1 5 4 3 7 8 7 0 1 7 2
2 5 9 0 5 5 1 9 6 2 1
3 6 0 1 7 0 1 5 9 0 1
4 7 6 6 5 4 9 0 3 8 0
5 2 6 9 7 4 2 9 5 7 9
6 8 8 4 2 5 0 7 0 8 2
7 7 9 0 8 0 2 0 5 8 1
8 7 1 3 7 0 2 0 9 9 3
9 2 2 6 1 9 8 6 0 2 6
arr = np.where(np.argsort(df_input, axis=1) < 5 , -1, 1)
df_output = pd.DataFrame(arr, columns=df_input.columns)
print (df_output)
a b c d e f g h i j
0 1 -1 1 1 -1 -1 -1 1 -1 1
1 1 1 1 -1 -1 -1 -1 1 1 -1
2 -1 1 1 1 -1 -1 -1 1 -1 1
3 -1 -1 1 -1 1 1 1 -1 -1 1
4 1 1 1 -1 -1 -1 -1 -1 1 1
5 -1 1 -1 1 -1 -1 1 -1 1 1
6 1 1 -1 1 -1 -1 1 -1 -1 1
7 -1 -1 1 1 1 1 -1 -1 1 -1
8 -1 1 -1 1 -1 1 -1 -1 1 1
9 1 -1 -1 -1 1 -1 1 1 1 -1
o = df_input.sort_values(by=list(df_input.columns), ascending=True, na_position='first')
o[list(df_input.columns)[:5]] = -1
o[list(df_input.columns)[6:]] = 1
You can rank, then assign conditionally via numpy.where:
df[:] = np.where(df.rank(axis=1) > df.shape[1] / 2, 1, -1)
print(df)
A B
0 1 -1
1 -1 1
2 -1 1
3 1 -1
Note: this assumes duplicate values always get same rank.
