改进 python ARIMA 预测

Question

I'm working on an ARIMA model in python for the first time and the predictions don't really make any sense. These are my values .

With auto_ARIMA i got the p, d and q values of 1, 1, 2.

model = ARIMA(train, order=(1, 1, 2)
fitted = model.fit(disp=0)
print(fitted.summary())

So I create the model and get this summary. I then split them into a train and test set, so I can validate the prediction and do the forecast:

fc, se, conf = fitted.forecast(len(test), alpha=0.05)  # 95% conf
fc_series = pd.Series(fc, index=test.index)
lower_series = pd.Series(conf[:, 0], index=test.index)
upper_series = pd.Series(conf[:, 1], index=test.index)

And get this plot .

I tested with other p, d, q values and didn't get a better result. The forecast line is always straight, never goes up and down. Why is this? How can I improve the result? Let me know if you need more information or plots.

Answer 1

This is usually due to the lack of strongly pronounced data features like seasonality or trend. Then ARIMA cannot handle the future properly and falls back to the average of data values.

I had a similar issue with my data and was able to improve the result by splitting it into smaller chunks.