Analytics Scoring Module
The Spectrumâ„¢ Technology Platform Analytics Scoring Module provides data preparation steps
that are specific to model scoring stages such as Binning Lookup, Java Model Scoring,
and PMML Model Scoring. It allows predictive models defined in either QMML (Spectrum
Miner's proprietary model format) or PMML (industry
standard Predictive Model Markup Language) to be evaluated within dataflows and
additionally allows dataflows to write and retrieve data from Miner datasets that can be
consumed by the Miner predictive analytics software. It also enables the scoring and
prediction of new data using binning or Java-based models that were created using
Machine Learning Module stages such as Logistic Regression, K-Means Clustering, Binning,
and so on. This in turn allows data to be enriched by adding predicted or scored outputs
using models created by data insight teams with industry-standard data modeling tools.
This may include models used to calculate the churn risk for existing customers or
credit scoring models to determine a consumer's credit rating.
Note: The Analytics
Scoring Module uses an underlying H2O.ai library for modeling algorithms in Java Model
Scoring.