Supported models
TimeSeries currently supports the following models:
| Model | Class | Description | One-step forecasting | Multi-step forecasting |
|---|---|---|---|---|
| AR(p) | AR | An autoregressive model of order p. | Fully tested. | Fully tested. |
| MA(q) | MA | A moving average model of order q. | Fully tested. | Fully tested. |
For more information on these models, see [1, 2, 3].
When forecasting using an nth order model and a forecasting horizon of h, the time and memory complexities of the models are as follows:
| Model | Time complexity | Memory complexity |
|---|---|---|
| AR | O(nh) | O(2n) |
| MA | O(nh) | O(3n) |
Note: these values were simply derived from the code, and are thus purely indicative. Computational experiments will be conducted in the future.
References
[1] George E.P. Box, Gwilym M. Jenkins, Gregory C. Reinsel, and Greta M. Ljung. Time series analysis: forecasting and control. John Wiley & Sons, 2015.
[2] William W.S. Wei. Time Series Analysis: Univariate and Multivariate Methods. Addison-Wesley, 2006.
[3] Aileen Nielsen. Practical time series analysis: Prediction with statistics and machine learning. O’Reilly Media, 2019.