Detecting the Future: All-at-Once Event Sequence Forecasting with Horizon Matching

TL;DR

The paper introduces DEF, a new method for simultaneous long-horizon event sequence forecasting that improves accuracy and diversity by aligning predictions with ground truth through a novel loss, outperforming existing models.

Contribution

It proposes a novel all-at-once forecasting approach with a matching-based loss, achieving state-of-the-art results in long-horizon event prediction.

Findings

Up to 50% improvement over existing models

State-of-the-art in next-event prediction

High computational efficiency during inference

Abstract

Long-horizon events forecasting is a crucial task across various domains, including retail, finance, healthcare, and social networks. Traditional models for event sequences often extend to forecasting on a horizon using an autoregressive (recursive) multi-step strategy, which has limited effectiveness due to typical convergence to constant or repetitive outputs. To address this limitation, we introduce DEF, a novel approach for simultaneous forecasting of multiple future events on a horizon with high accuracy and diversity. Our method optimally aligns predictions with ground truth events during training by using a novel matching-based loss function. We establish a new state-of-the-art in long-horizon event prediction, achieving up to a 50% relative improvement over existing temporal point processes and event prediction models. Furthermore, we achieve state-of-the-art performance in next-event prediction tasks while demonstrating high computational efficiency during inference.