- Explainability is a component of the Amazon Ensemble and MVLR algorithms. Explainability details pinpoint the relative impact of the forecast inputs (such as related data) on prediction results. Take, for example, an electric bike company:
- You can load a PlanIQ forecast model with information on historical oil prices, e-bike prices, and promotion data.
- Explainability could show whether the increased oil prices or promotions have a greater sales impact.
- Explainability details are derived from independent feature variables within algorithms. These details help to make predictions about the future. Some common feature examples are:
- Marketing campaigns and promotions
- Inventory levels
- Crude oil prices
- Not all features impact prediction results in the same way. The feature contribution provides an idea about the degree of influence a single feature has on a prediction result.
The impact of features on forecast results
The chart below displays the impact of features on forecast results.
You can see that 12-month seasonality has a much greater impact than historical trend or 3-month seasonality, which have similar impacts.