SDM_ILI

Git repository to write up the paper for SDM

TODO:

Deadline: Sep 17, 2013 Have all parts except intro and except results (but write 6.1 and 6.2) and except fusion Use SIAM DM format

Forecasting a moving target: Ensemble models for ILI case count predictions

1. Introduction
   naren will write this at the end.
   Contributions:

   • Multiple time series with staggered arrival rates
   • Revisions to published values are possible
   • Surrogates, models, ensemble
   • lead time as a consideration

2. Preliminaries

   • invent a notation for lagging arrivals and revisions vs stable values
   • State the problem formally. Define RMS metric (either absolute or relative)
   • Need a nice big block diagram

3. Models

   • then, for each time t_i, describe a model that can predict t_i
   • write descriptions for each model, 1 in each subsection

4. Ensembling approaches

   1. Data level fusion (Prithwish)
      • talk about MI filtering etc.
   2. Model level fusion (Pejman)
      • sequential ensembling
      • simultaneous ensembling
   3. Sensor fusion approach (Prithwish)
      • Prithwish will dig out some papers

5. Forecasting a moving target

   1. Include Pejman's preliminary analysis here. Revise the PAHO value itself.
   2. Assigning variance to PAHO data and using them in the models. Use N_muetras (sample size correction). Put a simple threshold? Use confidence-based nearest neighbor matching. Use this other information to revise the prediction.

6. Experimental results

   1. Datasets
      describe countries in LA, data sources, etc. pre-processing
   2. Evaluation measures
      lead time
   3. experiments
      look at how I like my experimental section organized.

7. Related work

8. Future work

---

• simdemics