Setting the grouping confidence level

Override the default grouping confidence level in a training job submission.

About this task

By default, the grouping confidence level for training is 90%, which is the recommended setting.

However, this level can be overwritten during a manual training job submission.

  1. Access the dev-trainer pod.

    oc rsh $(oc get pods|grep dev-trainer|awk '{print $1}')
  2. To run a job that uses the default confidence level:

    curl -X POST --header 'Content-Type: application/json' --header 'Accept: application/json' --header 'X-TenantID:cfd95b7e-3bc7-4006-a4a8-a73a79c71255' -d '{"properties":{}}' 'http://localhost:8080/1.0/training/train/related-events-cp4aiops' ; echo -e '\n'
  3. To change the confidence level, add re.minConfidence and re.minCooccurrenceLikelihood to -d '{"properties”:{}}’ and include a value between 0.1 and 1.0. Specify the same value for re.minConfidence and re.minCooccurrenceLikelihood. A value of 1.0 represents a confidence level of 100%, which means that all events must exist in all co-occurrences for the occurrence to be considered valid. Each 0.1 reduction in value represents a 10% decrease in confidence level (0.9 = 90%, 0.8 = 80%, and so on).

    The following example initiates a training run with a 60% confidence level.

    curl -X POST --header 'Content-Type: application/json' --header 'Accept: application/json' --header 'X-TenantID:cfd95b7e-3bc7-4006-a4a8-a73a79c71255' -d '{"properties":{"re.minConfidence":0.6, "re.minCooccurrenceLikelihood":0.6}}' 'http://localhost:8080/1.0/training/train/related-events' ; echo -e '\n'


For this training run, the grouping confidence and co-occurrence likelihood levels were changed.