Walkthrough: Train an ML.NET Machine Learning algorithm

In this walkthrough, you will learn how to:

• Define a digital twin model with properties to monitor for anomalies.

• Build a data set for supervised learning.

• Use the ScaleOut Machine Learning Training Tool to train an algorithm.

This tutorial will define a digital twin model for tracking sensor readings to determine if a mechanical part is overheating. The sensor values will be temperature, friction, and rotations per minute (RPM). The algorithm requires a set of labeled training data. To create the training set, use historical data where the combinations of these three properties led to mechanical parts overheating or operating normally. Input this data to train a machine learning algorithm to predict if a new set of values are likely to indicate overheating. After training the machine learning algorithm, modify the digital twin model implementation to leverage the algorithm and automatically flag overheating anomalies received from new messages.

Note

The training dataset, trained algorithm, and digital twin model are available on the ScaleOut DigitalTwinSamples GitHub repository.

• Defining a Digital Twin Model
• Creating the Training Data Set
• Training the Algorithm Using the ScaleOut Machine Learning Training Tool