如何修复Python中的ARIMA MODEL错误

Question

I'm trying to create an ARIMA Model for my time series data. What can i do to my code for it to run smoothly?

I'm using statsmodels to create the ARIMA model in python but i'm getting error warnings

    indexedDataset_logscale.head(10)
    OUTPUT: 
                Price
    Period  
    2013-02-08  2.515274
    2013-02-11  2.526528
    2013-02-12  2.520113
    2013-02-13  2.515274
    2013-02-14  2.543961
    2013-02-15  2.544040
    2013-02-19  2.530119
    2013-02-20  2.516082
    2013-02-21  2.508786
    2013-02-22  2.5273

    #AR Model
    from statsmodels.tsa.arima_model import ARIMA

    model = ARIMA(indexedDataset_logscale, order=(0, 1, 2))
    results_AR = model.fit(disp = -1)
    plt.plot(datasetLogDiffShifting)
    plt.plot(results_AR.fittedvalues, color = 'red')
    plt.title('RSS: %.4f' %sum((results_AR.fittedvalues-datasetLogDiffShifting['Price'])**2))
    print('Plotting AR Model')

Error messages i get are: 

• "ValueWarning: A date index has been provided, but it has no associated frequency information and so will be ignored when eg forecasting. ignored when eg forecasting.', ValueWarning)

• 8 plt.title('RSS: %.4f' %sum((results_AR.fittedvalues-datasetLogDiffShifting['Price'])**2)) TypeError: 'str' object is not callable

Answer 1

1) It seems your index doesn't have set frequency Try adding this befor Arima

df.index.freq = 'D'

D is for daily timestamp, it seems your is not linear/regular, hard to define looking at what's in the question. Check other frequency option in pandas.

To avoid warnings type this at the beginning of the script:

#Clear console
import warnings
warnings.filterwarnings("ignore")

Answer 2

The problem is that in your data, the index seems to be daily, but some dates are not consecutive days. You can supply missing dates and then interpolate values for these dates, as follows:

df = df.resample('D').mean()
df["Price"] = df["Price"].interpolate(method='linear', axis=0).ffill().bfill()

You will then be able to build a model and plot the fitted values.