Evaluation metrics has 2 main concepts: the metric set and the metric. Each metric is essentially a question that you would want to ask while a metric set is a grouping of metrics that you would specify when creating a model evaluation run.

We currently support 2 types rating and ranking metrics. Rating metrics will let you ask questions that can either be from 1 to 5 or 1 to 7. A rating metric will be asked for each model specified in the run.

Ranking metrics will let you rank each provided model in order of which is best to worst. Ranking metrics will apply globally to the task.

Return to: Model Evaluation