列必须与键长度相同

Question

我有这个Imputed DataFrame：

   Imputed_Df.head():


                        Atmospheric_Pressure    Global_Radiation        Net_Radiation   Precipitation   Relative_Humidity   Temperature    Wind_Direction   Wind_Speed
 Time                               
 2013-11-01 01:00:00             999.451              207.75               99.09            4.450000           39.958667         13.600000      117.231667    2.138500
 2013-11-01 05:00:00             992.760              167.77               85.16            5.746667           56.107500         11.900000      244.410000    2.313000
 2013-11-01 09:00:00             990.272              157.00               95.04            6.271000           37.113333         12.802083      297.131500    3.270350
 2013-11-01 10:00:00             998.367              191.26               82.32            4.428000           37.946500         13.800000      143.103333    2.232500

而我想要做的基本上是平滑所有列，而不是将新的平滑列添加到此DataFrame，所以这是我如何尝试这样做：

import statsmodels
from statsmodels.tsa.api import ExponentialSmoothing, SimpleExpSmoothing, Holt


def Smoothing(Col):
  for Col in Imputed_Df.columns:
    fit = SimpleExpSmoothing(Imputed_Df[Col]).fit(smoothing_level=0.2, optimized=False)
    fcast = fit.predict(start=Imputed_Df.index.min(), end=Imputed_Df.index.max())
    return fcast



 Imputed_Df[['col1', 'col2', 'col3','col4','col5' , 'col6' , 'col7' , 'col8']] = Imputed_Df.apply(Smoothing, axis=1)

但我得到了这个错误：

 Columns must be same length as key

任何建议都会得到很多赞赏，谢谢你。

Answer 1

假设你的Smoothing Function是Correct

print(Imputed_Df.apply(Smoothing, axis=1))

并检查你的count返回列的df它应该匹配8，因为你正在服用Imputed_Df[['col1', 'col2', 'col3','col4','col5' , 'col6' , 'col7' , 'col8']]

如果输出df列数不是8则尝试

import statsmodels
from statsmodels.tsa.holtwinters import SimpleExpSmoothing
def Smoothing(Imputed_Df):
    my_df = Imputed_Df.copy()
    for Col in Imputed_Df.columns:
        fit = SimpleExpSmoothing(Imputed_Df[Col]).fit(smoothing_level=0.2, optimized=False)
        my_df[Col] = fit.predict(start=Imputed_Df.index.min(), end=Imputed_Df.index.max())
    return my_df

其实我发现时间序列数据是不规则的。

Imputed_Df = Imputed_Df.resample('H').pad() ##
Imputed_Df[['col1', 'col2', 'col3','col4','col5' , 'col6' , 'col7' , 'col8']] = Smoothing(Imputed_Df)

我更喜欢这样写作

Imputed_Df[Imputed_Df.columns + "_SES"]= Smoothing(Imputed_Df)