[fix]: bug DateUtil + optimization #85
Conversation
|
Hello, can you describe me those changes ? |
metrics/dateUtil.py
Outdated
| df = df_anon.copy() | ||
| df_original_reduced = df_original.copy() |
There was a problem hiding this comment.
You have to keep this line because by not doing a copy, the data frames will be modified (even outside function)
There was a problem hiding this comment.
You have to keep this line because by not doing a copy, the data frames will be modified (even outside function)
Ok. I didn't know that. Thanks.
Hello, can you describe me those changes ?
Yeah, sure!
First, I keep the code which verifies weeks are all the same. Then because of that, we can retrieve the day_of_week using pandas.Series.dt.dayofweek .
abs(df_anon.date.dt.dayofweek - df_original.date.dt.dayofweek)
I calculate the difference between 2 series. The result is a pandas.Series whose values are from 0 to 6.
Now, I store 3 - result to df_score. 3 means 3 points (I'll divide later).
df_score = 3 - abs(df_anon.date.dt.dayofweek - df_original.date.dt.dayofweek)
Values of df_score is in [ -3 : 3]
return df_score.loc[df_score > 0].sum() / 3 / nb_orig_line
Finally, we choose only values that bigger than 0 and do the addition, we divide the result by 3 and number_of_lines.
Can you review my pull request? @danymat @beng-git