Result Explanation

PrCore has four default plugins. Here we explain the prescription output schema.

The prescription is a JSON object with the following fields:

• date: The timestamp of the prescription.
• type: The type of the prescription. Can be one of the following:
  • NEXT_ACTIVITY: Predicts the next activity of a case.
  • ALARM: Provides the probability of a negative outcome.
  • TREATMENT_EFFECT: Provides the treatment effect of a case.
  • RESOURCE_ALLOCATION: Provides the resource allocation recommendation of a case.
  • Others, depending on the plugins installed.
• output field is different for each plugin, and can be null if the plugin is not able to make a prediction.
• plugin is an object with the following fields:
  • name: The name of the plugin.
  • model: The name of the model used by the plugin.
  • accuracy, precision, recall, f1_score: Some plugins will return the evaluation metrics of the model.
  • Others, depending on the plugins installed.

This plugin predicts the next activity of a case. The output is a string, which is the name of the next activity.

{
    "date": "2023-03-04T12:31:53.670496",
    "type": "NEXT_ACTIVITY",
    "output": "A_ACTIVATED",
    "plugin": {
        "name": "KNN next activity prediction",
        "model": "count-encoding",
        "accuracy": 0.8446,
        "precision": 0.8523,
        "recall": 0.8446,
        "f1_score": 0.8414
    }
}

The output is the probability of a negative outcome.

{
    "date": "2023-03-04T12:31:53.626402",
    "type": "ALARM",
    "output": 0.364,
    "plugin": {
        "name": "Random forest negative outcome probability",
        "model": "count-encoding",
        "accuracy": 0.5472,
        "precision": 0.5491,
        "recall": 0.5472,
        "f1_score": 0.5481
    }
}

The output is an object with the following fields:

• proba_if_treated: The probability of a positive outcome if the case is treated.
• proba_if_untreated: The probability of a positive outcome if the case is not treated.
• cate: The Conditional Average Treatment Effect (CATE) score of the case.
• treatment: The treatment definition of the case. This is directly from the user’s previously inputted treatment definition.

{
    "date": "2023-03-04T12:31:54.638476",
    "type": "TREATMENT_EFFECT",
    "output": {
        "proba_if_treated": 0.6827,
        "proba_if_untreated": 0.0001,
        "cate": 0.6826,
        "treatment": [
            [
                {
                    "value": "O_SENT_BACK",
                    "column": "Activity",
                    "operator": "EQUAL"
                }
            ]
        ]
    },
    "plugin": {
        "name": "CasualLift treatment effect",
        "model": "count-encoding"
    }
}

The output is an object with the following fields:

• resource: The allocated resource.
• allocated_until: The timestamp until which the resource will be released. Which is the allocated timestamp plus the treatment duration.

Please note that this prescription is only provided under the streaming mode.

{
    "date": "2023-03-15T12:49:17.742267",
    "type": "RESOURCE_ALLOCATION",
    "output": {
        "cate": 0.6478,
        "resource": "Resource_B",
        "allocated_until": "2023-03-15T13:49:17.742254"
    },
    "plugin": {
        "name": "CasualLift resource allocation",
        "model": "SIMPLE_INDEX-length-3"
    }
}