Retrospective on Neural Networks

Neural Network models gained popularity in energy forecasting in the 1990’s. Unlike regression methods, neural networks are capable of estimating the shape of nonlinear relationships and the strength of interactions among variables. In this Brown Bag, we dig into this capability, compare the predictive power of neural networks to a list of regression methods and decision tree methods, and show how neural network derivatives can be used as a feature engineering tool for structured regression models.