Playground: An exclusive test environment to analyze Field Prediction accuracy

Playground: An exclusive test environment to analyze Field Prediction accuracy

Hello everyone,

We are introducing Playground: an exclusive proactive testing environment for businesses to analyze Zia's prediction accuracy. 

This allows admins and agents to:
  • Run predictions on actual tickets or specific content for realistic test conditions.
  • View predicted values along with a confidence score to assess accuracy.
  • Identify and correct failed predictions by annotating data.
  • Retrain Zia to improve future prediction performance by annotating. 
The Playground gives a first-hand opportunity to see how Zia's predictive ability fares. Agent can select a ticket or copy and paste email content to directly test how well Zia can predict the desired fields. Agent can review the prediction and fine-tune Zia as needed.

For example, if Zia must auto-classify requests by “Problem” as Display, Audio, or Battery based on the ticket content, the tickets can be analyzed immediately using playground. 



In case your business doesn't have tickets or doesn't keep record of certain kinds of requests, then you can simply copy-paste the desired content in the subject and description for Zia to analyze and give you a predicted value.

Playground gives instant analysis—in contrast to Sandbox.

If Zia does not predict a value or predicts incorrectly, it is indicated in red with a low confidence score. The confidence score is generated by analyzing the accuracy score of each picklist value. Businesses can take proactive measures to mitigate such cases in real-time prediction by annotating the correct field value and then retraining Zia.  

Over time, Zia's accuracy will improve and it will correctly classify similar tickets to help the team respond faster. 

Note:
  • Playground is available only for prediction models with accuracy scores above 0%.
  • If the accuracy score is less than the recommended confidence score, then the value is denoted in red, whereas if the accuracy score is higher than the recommended value, it is denoted in green. 
Notes
  • The reset option allows the agent to clear the prediction and restart with a different sample ticket or other content. 
  • While Playground's capabilities facilitate AI retraining to improve accuracy, the actual retraining process cannot be done in Playground

Playground is  available in US, IN, AU and EU DCs. For more details, please refer to the help doc.

PM: @Anonymous User 

Regards,
Raveena
User Ed | Zoho Des