Page tree
Skip to end of metadata
Go to start of metadata

What is PRC?

Probable Root Cause (PRC) is a machine learning process that identifies which Alerts are root causes of a Situation.

PRC looks for patterns in user supplied feedback. It does not use 'Root Cause Analysis' techniques. For a brief introduction watch the video below:

Why is PRC useful?

An operator needs to determine where they should start looking when they open a Situation to take a closer look at the Alerts. PRC provides pointers to where to begin troubleshooting and diagnosis so can speed up Situation resolution. If PRC is enabled, when the operator opens a Situation Room they will be shown the Top 3 Probable Root Cause Alerts under Next Steps tab.

How does PRC work?

Users manually label Alerts as either the a Root Cause Alert or a Symptom Alert, the AIOps PRC Model uses this data to predict Situation root causes.

Subsequently, when AIOps generates Situations, it labels an Alert or Alerts as having a Root Cause Estimate. A Root Cause Estimate is always assigned even if the data set is small. Generally, the more data AIOps has the more accurate it is. However, that data needs to be consistent and the model is only as effective as the data it is supplied with. For example, two conflicting labels will confuse the model. If you do not know the status of an Alert do not label it. You do not have to label every Alert.

How does AIOps learn?

Machine Learning uses features like Severity, Host, Description and Class and takes the values of those features for all labelled Alerts and uses a Neural Network to estimate the Root Cause for all the Alerts in a newly created Situation. It does this even if that Situation has not been seen before based on the model and labelled data. 

See PRC: Configure and Retrain for more information on training your model.

PRC Column

This column (Situation, Alerts Tab) shows the Probable Root Cause Estimate as a percentage of the Alerts in that Situation and is useful as a prioritisation aid. For example, the higher the value an Alert has, the higher the probability that the Alert is the root cause of the Situation

As Alerts are added to a Situation, the Root Cause is recalculated (Situation, Alerts list) and therefore the PRC column may change. The more accurate and consistent data you feed your model the more accurate the estimate.