I am bulding SARIMA time series with
statsmodels.tsa.statespace.sarimax beacuse pmdarima doesn’t install. My data has 44 observation 10 years every quarter. My target is to predict next 1 or 2 years. Could anyone give idea what I need to pot the prediction. I am not proficient in Python but I think there is kinf of missunderstanding between my quarterly data and the desired prediction. I compile algorityhm from towardsdatascience, articles from here and youtube.
After evaluating P,D,Q, m parameters with min AIC and fit the model this is the result – can’t plot the predict steps
I made 2 columns – dates and GVA – gross added value I am looking for Data set is here
If someone could help..
When the data is prepared (setting index right, stationarizing etc.), I usually do as follows:
model2 = sm.tsa.statespace.SARIMAX(df['x'], order=(0, 1, 3), seasonal_order=(0, 1, 1, 4)) res2 = model2.fit() pred_uc2 = res2.get_forecast(steps=12) # note here the usage of steps ahead and get_forecast pred_ci2 = pred_uc2.conf_int() ax = df['x'].plot(label = OB, figsize=(14, 8)) # test data pred_uc2.predicted_mean.plot(ax=ax, label=FC) $ prediction ax.fill_between(pred_ci2.index, # confidence intervals pred_ci2.iloc[:, 0], pred_ci2.iloc[:, 1], color='k', alpha=.25) ax.set_xlabel('Date') ax.set_ylabel('Price') plt.legend() plt.show()