CrossLag: Predicting Major Dengue Outbreaks with a Domain Knowledge Informed Transformer

2025-10-03

AI Summary

Accurate long-term forecasting of major dengue outbreaks for timely public health alerts remains challenging. This paper proposes a domain-knowledge-guided Transformer model, whose core innovation is the CrossLag attention mechanism: it explicitly models multi-scale temporal lags between external drivers—such as climate and oceanic anomalies—and dengue incidence via learnable lag encodings, enabling efficient fusion of internal and external time-series signals with minimal parameter overhead. Built upon the TimeXer framework, the model balances mechanistic interpretability with data-driven flexibility. Evaluated on real-world dengue incidence data from Singapore, it achieves statistically significant improvements over TimeXer and other baselines in 24-week horizon forecasting. Specifically, it enhances accuracy in both outbreak onset detection and peak intensity estimation. The proposed approach establishes a new paradigm for infectious disease early warning—robust, scalable, and inherently interpretable.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

A variety of models have been developed to forecast dengue cases to date. However, it remains a challenge to predict major dengue outbreaks that need timely public warnings the most. In this paper, we introduce CrossLag, an environmentally informed attention that allows for the incorporation of lagging endogenous signals behind the significant events in the exogenous data into the architecture of the transformer at low parameter counts. Outbreaks typically lag behind major changes in climate and oceanic anomalies. We use TimeXer, a recent general-purpose transformer distinguishing exogenous-endogenous inputs, as the baseline for this study. Our proposed model outperforms TimeXer by a considerable margin in detecting and predicting major outbreaks in Singapore dengue data over a 24-week prediction window.

Problem

• Predicting major dengue outbreaks requiring timely warnings
• Incorporating lagging environmental signals into transformer architecture
• Improving outbreak detection accuracy in Singapore dengue data

Innovation

• Domain knowledge informed transformer for outbreak prediction
• Incorporates lagging signals with low parameter counts
• Uses exogenous-endogenous inputs to detect outbreaks