Send Feedback

The final step of the utilizing TimeDetect is to send feedback for the predictions made using the POST /feedback endpoint

Providing feedback helps the model learn from real-world outcomes, and over time this will improve the accuracy and relevance of future predictions. You can do this by marking each prediction as either TRUE_POSITIVE, TRUE_NEGATIVE, FALSE_POSITIVE or FALSE_NEGATIVE.

Feedback type

If a registration is flagged as an anomaly but is not, it is a false positive. If the model fails to flag a registration that is anomalous, it is a false negative. If the model correctly flags an anomalous registration, it's a true positive. If it correctly ignores a normal registration, it's a true negative.

Feedback request

required

Array of objects

{

• "datasets": [

  • {

    • "datasetId": "52af4e6a-ee74-43a7-b5b6-c7508539a56b",
    • "jobId": "3a57dd18-299b-4129-a95f-00941c6f3909",
    • "predictions": [

      • {

        • "predictionId": "46878c8f-69fb-4727-83f3-694ffe28eda9",
        • "feedback": "FALSE_POSITIVE"

        }

      ]

    }

  ]

}

The feedback endpoint is designed to be used after the preidctions results has been reviewed - typically as a part of the approval process.