Industrial Internet Application Development

Determining outliers and offset management

In the real world, it is not always guaranteed that the models receive the good and expected datasets. There is always the possibility of human error or unexpected events that can produce the bad sets that can cause the models to produce bad predictions. It is always good practice to consider large datasets and identify the outliers during the process of building a model, and also to have a way of identifying the outliers in the incoming stream of data, which will typically be done as part of the model pipeline processing for the model predictions.