Configuring the model transparency policy for predictive models
By configuring the transparency threshold for a business issue and optionally adapting the transparency score for predictive model types, lead data scientists determine which predictive model types are compliant for that issue.
Non-compliant models might be forbidden to use by certain company policies. Each model type has a transparency score ranging from 1 to 5, where 1 means that the model is opaque, and 5 means that the model is transparent. Highly transparent model are easy to explain, whereas opaque models might be more powerful but difficult or not possible to explain. Depending on the company policy, models are marked as compliant or non-compliant. Model compliance is also included in the model reports that you can generate in the Prediction Studio.
In the navigation pane of Prediction Studio, click.
In the Transparency thresholds section, set the transparency threshold for each business issue.
The transparency threshold can be different for each business issue. For example, the Risk issue can have a higher threshold than the Sales issue. It means that models that are used for predicting risk need to be easy to explain.
In the Model transparency scores section, change the transparency score for individual model types.
- Predictive analytics
Predictive analytics predict customer behavior, such as the propensity of a customer to take up an offer or to cancel a subscription (churn), or the probability of a customer defaulting on a personal loan. Create predictive models in Prediction Studio by applying its machine learning capabilities or importing PMML models that were built in third-party tools.