df = pd.DataFrame({'year':[2019,2019,2019,2019,2020,2020,2020,2020,2021,2021,2021,2021],'month':[1,2,3,4,1,2,3,4,1,2,3,4],'values':[33,63,24,19,22,30,60,25,23,24,30,80]})

How can I add the percentage of change between each year/month. For example, if I compare 2019 to 2020 for the month 1 it would be:

22/30 = -33%

the expected result should show 0 for the first year (2019) and for the next years show the increase/decrease percentage-wise.

What I have tried is groupby transform but couldn't get any result.

CodePudding user response：

You can use a shift per year. For this increase the year by 1 and merge the DataFrame with itself. Then compute the percentage change:

cols = ['year', 'month']
vals = df[cols].merge(df.assign(year=df['year'] 1), on=cols, how='left')['values']

df['change'] = df['values'].sub(vals).div(vals).mul(100).fillna(0)

output:

    year  month  values      change
0   2019      1      33    0.000000
1   2019      2      63    0.000000
2   2019      3      24    0.000000
3   2019      4      19    0.000000
4   2020      1      22  -33.333333
5   2020      2      30  -52.380952
6   2020      3      60  150.000000
7   2020      4      25   31.578947
8   2021      1      23    4.545455
9   2021      2      24  -20.000000
10  2021      3      30  -50.000000
11  2021      4      80  220.000000

CodePudding user response：

If you reshape your df a bit you could use pd.pct_change()

df = df.set_index(['year', 'month']).unstack(level=0)
print(df)

      values          
year    2019 2020 2021
month                 
1         33   22   23
2         63   30   24
3         24   60   30
4         19   25   80


res = df.pct_change(axis='columns').mul(100).fillna(0)
print(res)

      values                        
year    2019        2020        2021
month                               
1        0.0  -33.333333    4.545455
2        0.0  -52.380952  -20.000000
3        0.0  150.000000  -50.000000
4        0.0   31.578947  220.000000