Hi, I am learning Quantitative Trading Strtegires and models Section 5.16. I noticed that the sample code using the close price from 2010-1-1 to 2020-2-1, and at the end it show the model performs well from 2019-2-1 to 2019-4-1. According to the videos given before, ARIMA model can predict the price in the future. How may I know the predicted outcome after 2020-2-1 like 2020-3-1?
In real world, if we could only get the previous 6-month data, may we predict the price in the next month?
Also. I used HangSeng Tech Index this year using this code, but the residual plot shows some outliers at left side.,which means ARIMA model may not work.Is there any alternative model to use? Thanks.
Hi Iris,
In the notebook, the ARIMA model was fitted on the data till 2019-01-29. Following this the, evaluate the model performance, the predictions were compared with data from 2019-2-1 to 2019-4-1.
To forecast further into the future, you can train the model with recent data, i.e. data till 2020-2-1 and can perform Rolling Window forecast as explained in the notebook.
This appends the input data of ARIMA model after each new prediction and predicts the price of next day. You can change the degree of rolling window according to the number of days you want to forecast.
You might have to use more than 6 months of data for building better model. Please train the model with more data of HangSeng Index and evaluate the model on test data. Please let us know if you need any assitance.
Thank you