两个数据框python的简单线性回归

Question

I have two dataframe with datetime index. the first dataframe may contain nan value while the second are not.

data1['A']
2019-06-01 00:00:00          NaN
2019-06-01 01:00:00          NaN
2019-06-01 02:00:00          NaN
2019-06-01 03:00:00          NaN
2019-06-01 04:00:00          NaN
                         ...    
2019-06-30 19:00:00    14.086600
2019-06-30 20:00:00    14.101033
2019-06-30 21:00:00    14.160733
2019-06-30 22:00:00    13.940633
2019-06-30 23:00:00    13.989567
Freq: H, Name: A, Length: 720, dtype: float64

data2['B']
2019-06-01 00:00:00    243.168989
2019-06-01 01:00:00    243.104673
2019-06-01 02:00:00    242.571222
2019-06-01 03:00:00    240.685214
2019-06-01 04:00:00    242.652392
                          ...    
2019-06-30 19:00:00    243.611821
2019-06-30 20:00:00    243.338931
2019-06-30 21:00:00    243.296361
2019-06-30 22:00:00    243.676107
2019-06-30 23:00:00    243.507886
Name: B, Length: 720, dtype: float64

how can I can conduct a simple linear regression model of those two dataframes when both have value at certain datetime (without nan). thanks for the help!

Answer 1

You can use LinearRegression from scikit-learn :

https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LinearRegression.html

Answer 2

您必须首先将数据框合并为一个数据框https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.merge.html ，然后在合并后的 scikit-learn 中尝试 LinearRegression数据框。

Answer 3

You can try something like this, you did not specify what to regress against what, so in the example below, I let x be 'A' and y be 'B':

from sklearn import linear_model
import pandas as pd
import numpy as np

data1 = pd.DataFrame({'A':[np.NaN,np.NaN,np.NaN,np.NaN,14.086600,14.101033,14.160733,13.940633,13.989567]})
data2 = pd.DataFrame ({ 'B':[243.168989,243.104673,242.571222,240.685214,242.652392,
                             243.611821,243.338931,243.296361,243.676107,243.507886]})
is_finite = np.isfinite(data1['A']) & np.isfinite(data2['B'])
mdl = linear_model.LinearRegression()
mdl.fit(data1.loc[is_finite][['A']],data2.loc[is_finite]['B'])
mdl.coef_