Define Settings and Train a Time Series Predictive Model
You enter values for parameters in the Settings tab. These are used to train a predictive model. Training is a process that takes these values and uses SAP machine learning algorithms to explore relationships in your data source to come up with the best combinations for the predictive model.
Enter new or change values for the following settings:
General section
- The predictive model name is a default one and can't be edited, but you can add descriptive text if required.
- Time Series Data Source: Browse to and select a dataset or a planning model that contains the historical data you want to use to train the predictive model. This data source must already be available in SAP Analytics Cloud.
- Version: For a planning model data source, select the version to use for training. It must be a public version, not in edit mode, or a private version. You have a least read access to it.
Predictive Goal section
- Target: Select the variable that you want to forecast values for.
- Date: Select the date variable. Check here for supported date formats: Restrictions.
- Time Granularity: The level of time granularity available in the data source.
- Number of Forecast Periods: Select the number of forecast periods that you want the predictive model to generate.
- Entity: Allows you to split a population into distinct entities. You select one or more nominal variables that identify each entity that you want to get forecasts for. For example you want to know sales forecasts for each country. A predictive model is generated for each entity that captures a specific behavior and produces distinct forecasts for each available combination of variable values. If this type of prediction is useful for you, then click the box, and select up to five variables to identify the entity that you want to split the population by.
Predictive Model Training section
- Train Using: The range of observations that will be considered for training. You can select all observations or define a window of time that the observations will be taken from.
- Until: The date of the last observation to be considered. You can select to use the last observation in the data set, or you can define a date.
Click Train and Forecastto start training the predictive model and generate the forecasts.
What's next?
Training and forecasting produces performance indicators that you will use to evaluate the results. This is called debriefing the predictive model.