如何通过dataframe的几列循环ARIMA

Question

I will start this by saying I am in no way a Python expert but my current project demands that it be programmed in Python, so any help and guidance is appreciated.

I have is a timeseries with daily data and 2000+ items.

I wish to run arima for each of these 2000+ columns. They are not dependent on each other. So basically it's like running 2000+ independent Arima analyses.

I have written a piece of code that loops through the columns and trains as per the parameters (order) provided to it. But it looks like as it trains further in the columns, it forgets what was learnt by the model before. Is there a way where I can make changes to the code where the trained results can be stored and used to the predict on the test set? Trying to predict for a couple of values in the last column. I am trying to train ARIMA for all the columns (Col1 to Col8 and predict couple of last values for Col 9)

Sample Dataset:

date                   Col1      Col2       Col3      Col4       Col5       Col6        Col7      Col8      Col9
2022-01-02 10:30:00     24         24        24.8      24.8       25         25         25.5      26.3      26.9   
2022-01-02 10:45:00     59         58         60       60.3       59.3       59.2       58.4      56.9      58.0   
2022-01-02 11:00:00     43.7       43.9       48        48        48.1       48.9       49        49.5      49.5   

#Test Train Split
train = df.iloc[:, :]
test = df.iloc[90:,-1]

order = (1,2,1) # <- plug-in p, d, q here 
for col in train.columns:
  model = ARIMA(train[col], order = order)
  model.initialize_approximate_diffuse()
  model = model.fit()
model.summary()

predictions = model.predict(len(test))

Answer 1

I think you have to do something instead of fitting the model inside the loop.

Let's know if this code works on your side

#Test Train Split
train = df.iloc[:, :]
test = df.iloc[90:,-1]

order = (1,2,1) # <- plug-in p, d, q here 

models=[]    # ---------> We create a list of different models here

for col in train.columns:
  model = ARIMA(train[col], order = order)
  model.initialize_approximate_diffuse()

  # model = model.fit() ----------> We replace this by :
  models.append(model.fit())

models[0].summary() #----------------> show first model summary

# Create a list of predictions

predictions=[]

for i in range(len(models)):
  predictions.append(models[i].predict(len(test)))