Analytics for the Internet of Things (IoT)

Feature engineering with IoT data

Beware of the siren call of the automated ML software tool, which takes all your data and determines what is important and automatically builds the best model from it–and all at the click of a button. Often, the raw data that you have is not in a form that ML models can be successful with. Using the data as it is can be a rocky proposition. Many an unaware ship has been wrecked on those rocks, lured by the lovely sound of automation.

One of the best ways to dramatically improve the predictive ability of your ML models is not in the algorithms themselves, but in how the data that they are grown from is presented to them. The transformations of data, the addition of constructed new fields, and the removal of ...