一つの方法は、boolean型のインデックスでdt.day
とdt.days_in_month
にある:
df = pd.DataFrame({'Date':pd.date_range('2010-01-01','2013-12-31',freq='D'),
'Value':np.random.rand(1461)})
df_out = df[df['Date'].dt.day > df['Date'].dt.days_in_month-5]
print(df_out.head(20))
出力:
Date Value
26 2010-01-27 0.097695
27 2010-01-28 0.236572
28 2010-01-29 0.910922
29 2010-01-30 0.777657
30 2010-01-31 0.943031
54 2010-02-24 0.217144
55 2010-02-25 0.970090
56 2010-02-26 0.658967
57 2010-02-27 0.189376
58 2010-02-28 0.229299
85 2010-03-27 0.986992
86 2010-03-28 0.980633
87 2010-03-29 0.258102
88 2010-03-30 0.827310
89 2010-03-31 0.813219
115 2010-04-26 0.135519
116 2010-04-27 0.263941
117 2010-04-28 0.120624
118 2010-04-29 0.993652
119 2010-04-30 0.901466