Sort:  

We have two sets of predictions. The first set is a blend made using traditional predictive models (anything supported by "DataRobot", which tests models across six+ different programming languages), while the AI predictions are made using Google's TensorFlow. Every predictive model is created and validated/cross-validated independently on a daily basis.

Wow, that seems so cool and complicated. Thank you for answering my question. I'll look into it!