Interpretable Feature Engineering for Time Series Predictors using Attention Networks

TL;DR

This paper introduces an interpretable feature engineering approach for time series regression using multi-head attention networks, combining convolutional layers and visualization tools to enhance interpretability and predictive performance.

Contribution

It develops a novel attention-based feature extraction method that explicitly captures temporal dynamics and improves interpretability in time series regression models.

Findings

Effective feature extraction demonstrated on real datasets

Improved interpretability through visualization tools

Good predictive performance achieved

Abstract

Regression problems with time-series predictors are common in banking and many other areas of application. In this paper, we use multi-head attention networks to develop interpretable features and use them to achieve good predictive performance. The customized attention layer explicitly uses multiplicative interactions and builds feature-engineering heads that capture temporal dynamics in a parsimonious manner. Convolutional layers are used to combine multivariate time series. We also discuss methods for handling static covariates in the modeling process. Visualization and explanation tools are used to interpret the results and explain the relationship between the inputs and the extracted features. Both simulation and real dataset are used to illustrate the usefulness of the methodology. Keyword: Attention heads, Deep neural networks, Interpretable feature engineering