Machine Learning Module

Introducing the Machine Learning Module

The Spectrum team is excited to share with you a preview of a powerful new module: Machine Learning. This technology preview is an early implementation of Machine Learning containing the fundamental capabilities that we think you’ll find most useful. We intend to add significant new capabilities in future releases, so to help make sure we are adding the features you want, we are making this technology preview available to you now. Your feedback will help guide the evolution of the Machine Learning Module. As you explore Machine Learning, keep this in mind:
  • This technology preview contains a limited set of features. When you find yourself saying "I wish I could…", let us know by filing an enhancement request with technical support. Your suggestions will help determine what features we add in future releases. For information about contacting technical support, see www.pitneybowes.com/us/contact-dcs.html.
  • Even the best software has some bugs. If you encounter a bug, let us know by submitting a bug report to technical support. Since this is a technology preview, we cannot guarantee that we will resolve your specific problem right away. For information about contacting technical support, see www.pitneybowes.com/us/contact-dcs.html.
  • Feel free to use this technology preview in your production environment. Keep in mind that we cannot adhere to normal service level agreements (SLAs) for technology previews.
  • We expect to get interesting and unanticipated feedback which could dramatically affect the next release of Machine Learning, so we cannot guarantee that you will be able to preserve all the work you do with Machine Learning when you upgrade to future releases.
  • Use your judgment when it comes to making business decisions based on insights generated with this technology preview. We cannot adhere to normal service level agreements (SLAs) for features that are in technology preview.

We hope you enjoy experimenting with the Machine Learning Module and look forward to hearing your feedback.

The Spectrum™ Technology Platform Machine Learning Module provides the ability to fit supervised and unsupervised machine learning models.
Note: The Machine Learning Module is supported only on Windows and Linux operating systems.

Binning

Binning divides records into groups (bins) for a continuous variable without taking into account objective information. You can perform unsupervised binning in one of two ways: using equal-width bins or equal-frequency bins.

K-Means Clustering

K-Means Clustering creates models based on analytical clustering, which segments a set of records into clusters of similar records based on data values.

Logistic Regression

Logistic Regression creates models from datasets that use binary objectives with input variables.

Java Model Scoring

This feature scores new data using the formula created when you fit a machine learning model.

Machine Learning Model Management

Machine Learning Model Management enables you to manage all machine learning models on your Spectrum™ Technology Platform server. You can expose, unexpose, or delete models. Additionally, you can view detailed information for each model and compare any two models of the same type.

Note: The Machine Learning Module uses an underlying H2O.ai library for modeling algorithms in K-Means Clustering, Logistic Regression, and Java Model Scoring.