CODEWITHSUNDEEP
pythonpython-3.xpandasnumpystatistics
Kshitij Yadav
Kshitij Yadav

Reputation: 1387

How to calculate percent change of each month, from average of previous months

I have data that looks like this:

YearMonth  PageViews  Users
202001      100        10
202002      150        12
202003      100        13
202004      120        15
202005      130        10

I want to find out the percentage difference of each new month from the average of the previous month's usage. For example,

  1. For 202001 it will be NA as there is no previous month
  2. For 202002 the average of the previous month will be only 202001, so the percentage difference for pageviews for 202002 will be 50% and user would be 20% as compared to 202001
  3. Similarly for 202003, the average of previous months would be average of 202001 and 202002 combined, so pageview average is 125 (100+150 divided by 2), and user is 11, so the percentage difference for 202003 will be -20% for pageviews and +18% for users and so on.

How can I find this using python? Any help would be appreciated. Thank you.

Upvotes: 0

Views: 257

Answers (2)

rhug123
rhug123

Reputation: 8768

Alternative using df.expanding().mean()

    ((df[['PageViews','Users']]
.div(df[['PageViews','Users']].expanding().mean().shift()))
.sub(1))

Output:

   PageViews     Users
0        NaN       NaN
1   0.500000  0.200000
2  -0.200000  0.181818
3   0.028571  0.285714
4   0.106383 -0.200000

Upvotes: 0

Quang Hoang
Quang Hoang

Reputation: 150745

Try cumsum()/range for the cumulative mean:

cols = ['PageViews','Users']
cum_mean = df[cols].cumsum().div(np.arange(len(df))+1, axis=0)
df[cols].div(cum_mean.shift()).sub(1)

Output:

   PageViews     Users
0        NaN       NaN
1   0.500000  0.200000
2  -0.200000  0.181818
3   0.028571  0.285714
4   0.106383 -0.200000

Upvotes: 2

Related Questions