Virtual Sensor analytics is a software sensor used to serve in place of a physical sensor for variables (called targets) that are too costly or impractical to measure.

It leverages machine learning techniques to learn the relationship between the target variable and a set of variables. Virtual Sensor analytics should be used when the target variable is supposed to have an underlying pattern and when historical measurements are available. These measurements could be data collected manually or by a temporary sensor.
Typical use cases: estimation of humidity on a production line in aggressive environment where no sensors can be used, detection of overcapacity on a crane…

Virtual Sensor analytics includes four features:

- Learn the model from past and target data (createModel).

- Apply the model on new data to estimate target (applyModel).

- Update an existing model with actual data (updateModel).

- Get information on an existing model