IDOBE: Infectious Disease Outbreak forecasting Benchmark Ecosystem
IDOBE: Infectious Disease Outbreak forecasting Benchmark Ecosystem
Authors: Aniruddha Adiga, Jingyuan Chou, Anshul Chiranth, Bryan Lewis, Ana I. Bento, Shaun Truelove, Geoffrey Fox, Madhav Marathe, Harry Hochheiser, Srini Venkatramanan Date: 2026-04-20 Paper ID: openalex:2604.18521
Summary
IDOBE is a novel benchmark ecosystem for infectious disease outbreak forecasting, addressing the lack of standardized evaluation frameworks in epidemiology. The repository aggregates over a century of surveillance data across multiple global locations, generating over 10,000 outbreak segments for multi-horizon short-term forecasting. The authors provide a comprehensive analysis of 11 baseline models, demonstrating that MLP-based approaches offer robust performance while statistical models perform competitively during pre-peak phases.
Key Contributions
- Introduces IDOBE, a standardized benchmark ecosystem containing over 10,000 curated epidemic outbreaks across 13 diseases to facilitate reproducible research.
- Evaluates 11 diverse baseline models (statistical vs. machine learning) across multiple horizons (1-4 weeks) using both point and probabilistic metrics (NWIS).
- Quantifies epidemiological diversity using information-theoretic and distributional measures to characterize dataset complexity for outbreak forecasting.
Open Questions & Future Work
Key Concepts
- idobe-benchmark-ecosystem: A standardized methodology for curating, segmenting, and benchmarking multi-horizon epidemiological outbreak time series.
Archivist Review
I have approved the IDOBE methodology as a reusable concept for epidemic time-series segmentation and benchmarking. I also approved the open question regarding the transition from univariate forecasting to multivariate, mechanistically-informed epidemic models. The dataset itself is a collection rather than a single, atomic repository, so I have excluded it from the dataset list to maintain the vault’s focus on high-impact, individual benchmarks.
Approved Concepts
- IDOBE Benchmark Ecosystem: Provides a standard, reproducible methodological framework for segmenting and evaluating epidemic outbreak time series, which is currently lacking in the field.
Approved Open Questions
- Multivariate Mechanistic Outbreak Forecasting: Bridging the gap between empirical univariate benchmarks and real-world, multivariate, and mechanistically-informed forecasting is a central challenge for future outbreak response.
Links
Metadata & Links
- url
- https://arxiv.org/abs/2604.18521
- paper_id
- 2604.18521
- paper_source
- openalex
- domain
- time-series
- tags
-
- architectures
-
- datasets
-
- concept_slugs
- idobe-benchmark-ecosystem
- dataset_slugs
-
- skill
- TimeSeriesSkill
- created_at
- 2026-04-23T05:47:07Z