Influenza Forecasting Framework based on Gaussian Processes
Authors: Christoph Zimmer, Reza Yaesoubi
ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We compare our framework to several state of the art benchmarks and show competitive performance. We, therefore, believe that our GP based framework for seasonal epidemics forecasting will play a key role for future influenza forecasting and, lead to further research in the area. |
| Researcher Affiliation | Collaboration | 1Bosch Center for Artificial Intelligence, Renningen, Germany 2Health Policy and Management, Yale School of Public Health, New Haven, USA. |
| Pseudocode | Yes | Algorithm 1 Seasonal Epidemics Forecasting |
| Open Source Code | No | The paper mentions using the GPML package but does not state that the code for their proposed framework is open-source or provide a link to it. |
| Open Datasets | Yes | We use the official CDC data on influenza-like illness (ILI) as data source (Centers for Disease Control and Prevention, a). |
| Dataset Splits | Yes | We use the seasons 2003/04 – 2007/08 as training data and the seasons 2010/11 and 2011/12 as validation data to determine appropriate feature sets I of previous weeks for the training data. |
| Hardware Specification | No | The paper does not specify any hardware used for running the experiments. |
| Software Dependencies | Yes | We use Matlab 2016a and the GPML package (Rasmussen & Nickisch, 2019) for training Gaussian processes and predictions. |
| Experiment Setup | Yes | We use the seasons 2003/04 – 2007/08 as training data and the seasons 2010/11 and 2011/12 as validation data to determine appropriate feature sets I of previous weeks for the training data. We use various sets I (with up to five past weeks as features, I {i − 4, i − 3, . . . , i }) and choose the best three performing sets to become part of the ensemble as stated in Algorithm 1. We perform 1-4 week forecasts and therefore, end up with a total of 7 seasons, 29 weeks per season and 4 targets is in the order of minutes with our framework. |