Traitement des données de séries temporelles dans les pandas

Question

J'ai des données de séries chronologiques quotidiennes. J'essaie de corriger ces valeurs de séries temporelles en multipliant certains facteurs mensuels. il l'a fait comme la correction manuelle comme:

mes données de séries chronologiques ressemble:

model: 

2010-01-30 0.008909 
2010-01-31 0.007562 
2010-02-01 0.
2010-02-02 0.010286 
2010-02-03 0.012244 
2010-02-04 0.011367 
2010-02-05 0.010800 
2010-02-06 0.007610 
2010-02-07 0.006534 
2010-02-08 0.004721 
       ... 
2015-12-02 0.005415 
2015-12-03 0.004358 
2015-12-04 0.006844 
2015-12-05 0.002373 

J'ai un des facteurs mensuels comme:

mon_slope: 

month 
January  -0.168627 
February -0.165102 
March  -0.112321 
April  -0.112232 
May   -0.080092 
June  -0.129905 
July  -0.078751 
August  -0.095756 
September -0.090188 
October  -0.109919 
November -0.155380 
December -0.137885 
Name: slope, dtype: float64 

ce que je faisais:

jan_corr = pd.DataFrame(model[model.index.month ==1]*mon_slope.ix[0][1]) 
feb_corr = pd.DataFrame(model[model.index.month ==2]*mon_slope.ix[1][1]) 
mar_corr = pd.DataFrame(model[model.index.month ==3]*mon_slope.ix[2][1]) 

.................. 
.................. 

final = pd.concat([jan_corr,feb_corr,mar_corr]) 

mais je suis sûr que ce n'est pas la bonne façon de faire. est-il des moyens plus simples de le faire:

Answer 1

D'abord, créer un mappage:

mapping = dict(months.values) 
mapping 

{'April': -0.112232, 
'August': -0.095756, 
'December': -0.13788499999999998, 
'February': -0.165102, 
'January': -0.168627, 
'July': -0.078751, 
'June': -0.129905, 
'March': -0.112321, 
'May': -0.080092, 
'November': -0.15538, 
'October': -0.109919, 
'September': -0.090188} 

Vous pouvez utiliser Series.dt.strftime pour récupérer MonthNames:

df.iloc[:, 0].dt.strftime('%B') 
Out[143]: 
0  January 
1  January 
2  February 
3  February 
4  February 
5  February 
6  February 
7  February 
8  February 
9  February 
10 December 
11 December 
12 December 
13 December 
Name: 0, dtype: object 

Maintenant, l'utiliser pour accéder aux multiplicateurs df.replace ou df.map:

df.iloc[:, 1] = df.iloc[:, 0].dt.strftime('%B').map(mapping) * df.iloc[:, 1] 
df 

      0   1 
0 2010-01-30 -0.001502 
1 2010-01-31 -0.001275 
2 2010-02-01 -0.002043 
3 2010-02-02 -0.001698 
4 2010-02-03 -0.002022 
5 2010-02-04 -0.001877 
6 2010-02-05 -0.001783 
7 2010-02-06 -0.001256 
8 2010-02-07 -0.001079 
9 2010-02-08 -0.000779 
10 2015-12-02 -0.000747 
11 2015-12-03 -0.000601 
12 2015-12-04 -0.000944 
13 2015-12-05 -0.000327 

---

Détails:

df 

      0   1 
0 2010-01-30 0.008909 
1 2010-01-31 0.007562 
2 2010-02-01 0.
3 2010-02-02 0.010286 
4 2010-02-03 0.012244 
5 2010-02-04 0.011367 
6 2010-02-05 0.010800 
7 2010-02-06 0.007610 
8 2010-02-07 0.006534 
9 2010-02-08 0.004721 
10 2015-12-02 0.005415 
11 2015-12-03 0.004358 
12 2015-12-04 0.006844 
13 2015-12-05 0.002373 

months 

      0   1 
0  January -0.168627 
1 February -0.165102 
2  March -0.112321 
3  April -0.112232 
4   May -0.080092 
5  June -0.129905 
6  July -0.078751 
7  August -0.095756 
8 September -0.090188 
9  October -0.109919 
10 November -0.155380 
11 December -0.137885 

---