如何从元组列表中生成 dataframe，其中元组是值？

Question

I have a time series dataset on which I am running the auto arima model. The dataset has multiple columns that are independent of each other, so basically it's like multiple auto arima analysis.

The code I currently have loops through all the columns in the dataframe and stores the order values of p,d,q for each column in a list. What I want to achieve is: to store the p,d,q values for each column in a dataframe row wise.

Time Series Dataframe

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 

Code

##Auto arima
# def arimamodel(series):
autoarima_results=[]  
series = df.columns
for col in series:
    print("Auto Arima for : ", {col})
    ARIMA_model = pm.auto_arima(
        df[col],
        start_p=1,
        start_q=1,
        test="adf",
        max_p=5,
        max_q=5,
        d=None,
        trace=True,
        error_action="ignore",
        suppress_warnings=True,
        stepwise=True,
    )
    ARIMA_model.summary()
    autoarima_results.append(ARIMA_model.order)

This returns a list that looks like: [(1,1,0), (2,1,1), (1,1,1)]

For example, the orders of p,d,q suggested by auto arima are, Col1: 1,1,0, Col2: 2,1,1, Col3: 1,1,1 and so on.

The final output should be a dataframe that would look like is as below. Where every row represents one column and its p,d,q values:

Results         pdq_values
Col1            (1,1,0)      
Col2            (2,1,1)    
Col3            (1,1,1)

Answer 1

Let's say that:

• df has three columns: "Col1", "Col2", "Col3"
• and autoarima_results == [(1, 1, 0), (2, 1, 1), (1, 1, 1)]

Then, here is one way to do it:

new_df = (
    pd.DataFrame(autoarima_results, index=cols)
    .pipe(lambda df_: df_.assign(pdq_values=df_.apply(lambda x: tuple(x), axis=1)))[
        "pdq_values"
    ]
    .to_frame("pdq_values")
)

new_df.index.name = "Results"

print(new_df)
# Output
        pdq_values
Results
Col1     (1, 1, 0)
Col2     (2, 1, 1)
Col3     (1, 1, 1)

Answer 2

Starting with the output of your ARIMA model, here is a simple way to do this without using lambda functions -

#Your ARIMA output
l = [(1,1,0), (2,1,1), (1,1,1)]

#Convert the list of tuples into entries from a single column
df = pd.DataFrame({'pdq_values':l})    #<----
#df = pd.DataFrame([l]).T              #Another way!

#Change the index by adding 1, setting column name and prepending 'Col'
df.index = 'Col' + (df.index.set_names('Results')+1).astype(str)

#Reset index to get the index as a column in the df
df = df.reset_index()

df

  Results pdq_values
0    Col1  (1, 1, 0)
1    Col2  (2, 1, 1)
2    Col3  (1, 1, 1)

Additional notes:

1. The trick is to pass it as a dictionary (key, value pair) where you can specify the column name as key {'pdq_values':l} . Pandas reads this as a single column where each tuple is the entries for each row respectively. Another way to force this behavior is to pass the list of tuples as a list of lists [l] . This would create a dataframe with n columns and then you will need to do a transpose.
2. 'Col' + (df.index.set_names('Results')+1).astype(str) does 4 things at once. Changes the name of the index column to Results , adds 1 to it, converts it to a string, and prepends it with Col . This results in 0,1,2.. to become Col1, Col2, Col3..