DMIDAS: Deep Mixed Data Sampling Regression for Long Multi-Horizon Time Series Forecasting

TL;DR

DMIDAS introduces a novel neural forecasting method that enhances long-horizon predictions by combining smoothness regularization and mixed data sampling, achieving higher accuracy and efficiency on healthcare and electricity data.

Contribution

The paper proposes DMIDAS, a new approach that improves long-term time series forecasting accuracy and reduces model complexity using regularization and data sampling techniques.

Findings

Achieves 5% better accuracy than state-of-the-art models on long-horizon data.

Reduces NBEATS parameters by nearly 70%.

Effective on high-frequency healthcare and electricity price data.

Abstract

Neural forecasting has shown significant improvements in the accuracy of large-scale systems, yet predicting extremely long horizons remains a challenging task. Two common problems are the volatility of the predictions and their computational complexity; we addressed them by incorporating smoothness regularization and mixed data sampling techniques to a well-performing multi-layer perceptron based architecture (NBEATS). We validate our proposed method, DMIDAS, on high-frequency healthcare and electricity price data with long forecasting horizons (~1000 timestamps) where we improve the prediction accuracy by 5% over state-of-the-art models, reducing the number of parameters of NBEATS by nearly 70%.