Interpretowalne prognozowanie za pomocą sieci neuronowych
[Interpretable neural forecasting]

Kin G. Olivares and Cristian Challu 
(Machine Learning Department, Carnegie Mellon University, Pittsburgh, USA)


We discuss two neural network architectures that have an interpretable configuration 
and are capable of visualizing the relative impact of trend and seasonal components 
and revealing the modeled processes' interactions with exogenous factors.

In the first part of the talk we extend the neural basis expansion analysis (NBEATS) 
to incorporate exogenous factors. To showcase the utility of the resulting NBEATSx 
model, we conduct a comprehensive study of its application to electricity price 
forecasting tasks across a broad range of years and markets. We observe 
state-of-the-art performance, significantly improving the forecast accuracy by 
nearly 20% over the original NBEATS model, and by up to 5% over other well 
established statistical and machine learning methods specialized for these tasks. 

In the second part, we incorporate smoothness regularization and mixed data sampling 
techniques to the NBEATS architecture. We validate the resulting model - Deep Mixed 
Data Sampling Regression (DMIDAS) - on high-frequency healthcare and electricity price 
data with long forecasting horizon. We improve the prediction accuracy by 5% over 
state-of-the-art models, reducing the number of parameters of NBEATS by nearly 70%.