如何根据另一列的两个连续值在 pandas 的新列中添加 label？

Question

I've got a dataframe, df , with a single column, extension .
The values in extension column are cyclically increasing and decreasing like below:

extension
0.000
0.050
0.100
0.150
0.130
0.080
0.020
0.050
0.075

I'm trying to label each increasing and decreasing cycle like below:

extension lablel
0.000      1
0.050      1
0.100      1
0.150      1
0.130      1
0.080      1
0.020      1
0.050      2
0.075      2

I'm a bit stuck, and would appreciate some guidance here.

Answer 1

df['lablel']=df.extension.diff()#Find the difference between consecutive ros in the column extension
df['lablel']=(df.lablel.ge(0)&df.lablel.shift(1).le(0)|df.lablel.ge(0)&df.lablel.shift(-1).le(0)).cumsum()+1#Find zero crossing from the consecutive differences, cummulatively sum and add 1 to the outcome



 extension  lablel
0      0.000       1
1      0.050       1
2      0.100       1
3      0.150       2
4      0.130       2
5      0.080       2
6      0.020       2
7      0.050       3
8      0.075       3

Answer 2

So lets reproduce your data:

a = [0.000,0.050,0.100,0.150,0.130,0.080,0.020,0.050,0.075]
df = pd.DataFrame(a, columns=["extension"])

The short answer is this:

df["label"] = pd.Series(np.where(df["extension"].diff() < 0, 0, 1)).diff().abs().cumsum() + 1
df.at[0,"label"] = 1

At least that's my answer. But it definitly looks a bit clunky. So let's break it down step by step for understanding:

df["extension"].diff()

diff creates the difference between each cell and the previous. Therefore it cannot calculate it for the first element.

Output:

0      NaN
1    0.050
2    0.050
3    0.050
4   -0.020
5   -0.050
6   -0.060
7    0.030
8    0.025

Now let's binarize the result to detect changes in positive/negative difference, using where from numpy:

np.where(df["extension"].diff() < 0, 0, 1)

Output:

array([1, 1, 1, 1, 0, 0, 0, 1, 1])

This tells us if difference to previous is negative (--> 0) or positive (--> 1)

Then you want to know only when the positive/negative trend changes. Therefore we incorporate the diff function once more. Beforehand we have to convert the numpy array back to a pd.Series :

pd.Series(np.where(df["extension"].diff() < 0, 0, 1)).diff()

Output:

0    NaN
1    0.0
2    0.0
3    0.0
4   -1.0
5    0.0
6    0.0
7    1.0
8    0.0

Ultimately you're not interested in which direction the trend has changed, only THAT it changed, therefore we erase this information with the abs function. And then sum the result up with the cumsum function so that it can increase on every change:

pd.Series(np.where(df["extension"].diff() < 0, 0, 1)).diff().abs().cumsum()

Output:

0    NaN
1    0.0
2    0.0
3    0.0
4    1.0
5    1.0
6    1.0
7    2.0
8    2.0

Finally two additions to base the label at 1 rather than 0 and to replace the first item that was NaN: +1 behind the code and df.at[0,"label"] = 1

And there you go:

         extension  label
    0      0.000    1.0
    1      0.050    1.0
    2      0.100    1.0
    3      0.150    1.0
    4      0.130    2.0
    5      0.080    2.0
    6      0.020    2.0
    7      0.050    3.0
    8      0.075    3.0

EDIT: answer to edited question in the comments