Quantile Regression In Large Scale Energy Datasets

        In most of the prediction problems, we are often interested in knowing the uncertainty in our predictions. Knowing about the range of predictions, as opposed to only knowing the point estimates, can significantly improve decision making processes for many problems. Quantile regression helps to provide sensible prediction intervals even for datasets whose residuals have non-constant variance or non-normal distribution.

        Quantile Regression (QR) problems can be formulated as Linear Programming (LP) problems. However, these LPs turn out to have large size as the size of the dataset could be very large. By using Random Projection (RP), they could be solved approximately but efficiently.