EpiCastBench: New Benchmark for Multivariate Epidemic Forecasting
EpiCastBench, a novel benchmarking framework, has been launched to fill the gap in high-quality benchmarks for multivariate epidemic forecasting. It includes 40 carefully selected multivariate epidemic datasets, which are publicly accessible and cover a variety of infectious diseases. These datasets vary in temporal granularity, length, and sparsity. The initiative responds to the growing trend of data-driven decision-making in public health, as multivariate forecasting models can more effectively capture intricate temporal relationships compared to traditional univariate methods. The datasets undergo analysis to uncover global characteristics and structural attributes. This framework is designed to support the creation of effective epidemic forecasting techniques through a standardized evaluation platform.
Key facts
- EpiCastBench is a large-scale benchmarking framework for multivariate epidemic forecasting.
- It features 40 curated multivariate epidemic datasets.
- Datasets are publicly available and span a wide range of infectious diseases.
- Datasets exhibit diverse temporal granularity, series length, and sparsity.
- Multivariate models capture complex temporal dependencies better than univariate approaches.
- The framework addresses the lack of high-quality benchmarks in epidemic forecasting.
- Datasets are analyzed for global features and structural properties.
- The work supports data-driven decision-making in public health.
Entities
—