Models and Scores

Darwinium provides score attributes. 

A Model Score can be a better way to evaluate risk, and apply subsequent decision thresholds compared to individual Rules alone.

Types of  Scores and Models

Scores are contained inside the map attribute and accessed with:

• step_name

• model_name (given in the Workflow of that step in journey file)

outcome[CHAMPION].models.score['step_name.model_name']

An easier expression for accessing this attribute, independent of step:

modelScore('model_name')

Ruleset Score

Darwinium Rulesets also can function as models. The ruleset score is the linear sum of all individual scores for rules that evaluated to True. The score takes an integer value between -1000 (most risky) to +1000 (least risky), where more or less is clipped to those values.

The scores and rule score contributions are shown in the Event View to aide investigators.

Machine Learning Models

Machine learning models can be executed in the Darwinium engine, outputting a float value (eg. score, probability).

Machine Learning Execution

Decision Strategy Score

Decision strategy is a special ruleset which always runs last. They have a single score that works in same way as a ruleset and  is output to a separate attribute.

outcome[CHAMPION].decision_strategy.score

Using Scores in Decisions

The Model Scores and Results can be referenced in Rulesets, including the final Decision Strategy to contribute to decisions.

modelScore('model_name') <= -800