For the list of all supported Decision Insight downloads and releases, see the Downloads
Classifiers and thresholds for evaluating a situation
Currently, your dashboard is very basic. It displays the count and chart for three metrics:
- The number of orders received today.
- The number of orders completed today.
- The number of orders currently being processed.
Part 4 of the tutorial focuses on adding additional information to your dashboard so you can know:
- Whether the numbers you're seeing are within expectations.
- Whether a situation is improving or getting worse.
In other words, let's add more contextual information that will enable you to better evaluate the risks in a situation.
Classifiers define changes of state and the visual clues to provide the user to identify those changes. Classifiers are:
- The building blocks of status evaluations. You use classifiers to define how evaluations will change from one state to the other.
- Defined at the deployment level – one classifier may be used for more than one attribute.
Thresholds are associated with an entity, just like attributes and relation attributes.
This makes sense because entities can be quite different.
For example, let's say you have an entity that tracks fuel levels and another that monitors speed. The values (and direction of status changes) would be different for each entity's threshold but they could share the same classifier.
In the figure below:
- Classifiers – define the colors of the risk levels and their corresponding labels – Normal, Warning, Danger.
- Thresholds – define the actual values of the attributes that will trigger a change in status
There is a difference in evaluating situations rather than simply reporting numbers. Evaluations provide an analysis of a situation that – without DI – might only come from years of experience.
Numbers provide facts. Thresholds and evaluations provide opinions about facts.
In your application, the next logical step is to learn how to define evaluations and thresholds.
First, you need to access more realistic data and expand your model to integrate additional entities. That comes next.