Embedded machine learning enhancements

This feature expands the Objective Model Types supported by the Embedded ML capabilities to include the Planned and Event based ETA prediction as well as expanding the Embedded Machine Learning models to include Scenario Filters and Scenario Outlier Filters.  With the delivery of these features - you now have the ability to run all three of the available Prediction Models end-to-end as an embedded wholly contained within the application. 

The Embedded ML capabilities for Planned and Planned and Event based ETA prediction provides you with the ability to use Machine Learning capabilities to predict the transit time of your OTM shipments.  You will find the Embedded Logistics Machine Learning capabilities to be extremely beneficial for predicting the transit time for all modes of transportation where the transit time calculation is based on a schedule and/or a static zone-to-zone or postal code to postal code lookup.

Specifically, you will find the Embedded Machine Learning prediction capabilities to be extremely beneficial for predicting the ETA for the following modes of transport:

  • Air - where the ETA is based on published flight schedules,
  • Ocean - where the transit time and ETA are based on published voyage schedules,
  • Rail – where the transit time and ETA are based on published rail schedules,
  • Intermodal - where, from a shipper’s perspective, the ETA is often based on a “slow truck” simulation guesstimate,
  • Barge – where the transit time maybe based on a published schedule, but is typically modeled as a simple lookup guesstimate,
  • Less-Than-Truckload – where the transit time is based on a published postal code to postal code lookup table,
  • Parcel – where the transit time and ETA are based on either a “guaranteed” level of service - for example, Overnight AM, Next Day etc, or the service is based on a static zone to zone lookup table to determine the transit time/transit days between two zones.

With the transition to the Embedded Machine model, all of the steps required for selecting the data to training your model, train the model and the perform predication is now embedded within the application.  At a high level the steps required include:

  1. Creating a Machine Learning Project
  2. Defining data to train - setup Machine Learning Scenario(s)
  3. Performing Training on the selected Machine Learning Scenarios
  4. View Training Results - Review Feature Importance Summary from training
  5. Running Perform Prediction on an OTM operational shipment
  6. Reviewing Prediction Results

Feature extends the innovative embedded machine learning capability provided natively within OTM for event based prediction.

Steps to Enable

You don't need to do anything to enable this feature.

Tips And Considerations

ML Training is a resource intensive project. Customers are advised to run training is off-peak hours.