Designing the time series data model
We are going to explain the general steps to building a time series model using training data. Use your judgment and experience to discern from the chart whether the data is autoregressive before applying the Durbin-Watson statistical test.
The sequential steps required to build the predictive model with time series machine learning are as follows:
- Plot the data to inspect the possible autocorrelation relationship.
- Use the Durbin-Watson statistical test to see whether the data is autocorrelated.
- Calculate the centered moving average of each period lag of the data.
- Determine the separation between the data and the centered moving average. This is known as seasonal irregularity.
- Get the trending component of the time series using the regression model line.
- Multiply the seasonal irregularity value by the trending result to make the forecast.
We use the centered moving average to smooth or to take the general...