Application of Gradient Boosting in Order Book Modeling

The basic metric of success is to get the error less than the baseline. It means that the final model has good quality. The first question is how to measure quality. It could be squared errors. After that, we can estimate the interval by bootstrapping method.