1

I have a some dataframe object:

df.head()

A   B   C
0   7   7     
1   3   6     
2   5   5  
3   2   7     
4   4   3

I would like to compare for example A and B at n while also using C at n-1 to create a new row, so let's say D = An * Bn + Cn-1.

At first I used a generic python loop, but I quickly realized this was very slow with large datasets. Then I started looking at numpy vectorization (which is very fast) but I couldn't figure out a way to get previous entries.

What other alternatives do I have while keeping it nice and fast?

Duco
  • 47
  • 7
  • 1
    First I think need something like [this](https://stackoverflow.com/a/62261221/2901002), if not `shift` is better here. – jezrael Aug 16 '20 at 10:49

1 Answers1

4

Another solution using Series.shift

df['D'] = df.A * df.B + df.C.shift(1, fill_value = 0)

   A  B  C   D
0  0  7  7   0
1  1  3  6  10
2  2  5  5  16
3  3  2  7  11
4  4  4  3  23
sushanth
  • 8,275
  • 3
  • 17
  • 28
  • 1
    Thank you. fillna(0.0) is redundant as shift() has a fill_value parameter, so shift(periods=1, fill_value=0.0). – Duco Aug 16 '20 at 10:40