ML for Energy: why academic models fail in the real world

14:43-14:55, January 27 @ 2BC

Talk/ Overview

Forecasting power and natural gas demand are classical problems in the scientific literature. However, real-world business requirements often introduce additional complexity: bad data quality, delays in sample availability, need for stratified predictions and inaccuracy in weather forecast may undermine the performance of predictors designed for ideal conditions. With real case studies, we show examples of such issues, and how proper feature engineering and modelling may offset the negative effects on overall prediction accuracy.

Talk/ Speakers

Emanuele Fabbiani

Co-Founder & Chief Data Scientist, xtream

Talk/ Highlights

10:45

ML for Energy: why academic models fail in the real world

With Emanuele FabbianiPublished March 12, 2020

AMLD / Global partners