How to run ARIMA model for each column of a dataframe?

Question

I´m predicting the prices for a good in 5 regions. The data is organized as a Pandas data frame.

When I run autoARIMA for each column apart, it is fine (eg, for data_frame["Region_name"] ).

def __trainArima(self, actual_values, periods):
     fitted_model = pm.auto_arima(actual_values, start_p=3, 
                                 start_q=2,
                                 max_p=3, max_q=3, m=12,
                                 start_P=0, seasonal=True,
                                 d=1, D=1, trace=True,
                                 error_action='ignore',
                                 suppress_warnings=True,  
                                 stepwise=True) 
     return fitted_model.predict(n_periods=periods)

__trainArima(data_frame, 12)

However, I want to run it at once for all the 5 columns, having an output as a data frame of 5 columns for each region of predictions for the next 12 month. Is that possible?

Answer 1

如果您不相信自己的列是相关的，则可以遍历每个列并运行ARIMA，然后合并预测。

Answer 2

To elaborate, you could loop though like this:

import pandas as pd

results = []
cols = []
for i in data_frame.columns:
    cols.append(i)
    result = __trainArima(test[i])
    results.append(result)

output = (pd.DataFrame(results)).T
output.columns=cols

that will output a data frame with your predictions under the same column names as they appear in the original dataframe.