- Darpa-chikun-challenge - mailman.cs.vt.edu

DARPA CHIKV Challenge Submission Confirmations
by Prithwish Chakraborty 01 Dec '14

01 Dec '14

Using this thread for confirmation of submissions. Dec 01 submissions done (confirmation attached) and submission can be found at https://copy.com/qTgIx7sqkvSc6ZI6 . I am working on a switching Dynamic Model. Will update you all soon. Thanks Prithwish Chakraborty Graduate Research Assistant Dept. of Computer Science Discovery Analytics Center Virginia Polytechnic Institute and State University email: prithwi(a)vt.edu web: http://people.cs.vt.edu/~prithwi/

1 0

CHIKV Challenge Submission #2
by Prithwish Chakraborty 23 Sep '14

23 Sep '14

Hi All, Our next submission is due Oct 1. Back at Arlington, Bryan, Brian and me came up with a few additions to the model mainly 1. Investigating the disease parameters more closely 2. Using a SEIR for base model 3. Constructing a Data driven SEIR to incorporate travel and related information (coding undergoing). Please see the update on that attached herewith. I was wondering whether we can meet sometime tomorrow to share the thoughts on this so that I can finish the coding and get results by Friday. I am proposing either 3-4pm Sep 24, 2014 (tomorrow) or 10-11am Sep 25, 2014 (day after tomorrow). Can you please share your preferences about the same in the following doodle link? http://doodle.com/s4bfat529xpa7qb4 (it has more time-slots) Sorry for being a little late on this update. Thanks Prithwish Chakraborty Graduate Research Assistant Dept. of Computer Science Discovery Analytics Center Virginia Polytechnic Institute and State University email: prithwi(a)vt.edu web: http://people.cs.vt.edu/~prithwi/

1 0

Fwd: DARPA challenge FAQ update
by Prithwish Chakraborty 01 Sep '14

01 Sep '14

Dear Solvers, We have provided these as a response to some common questions. Please note that we have also updated the provided spreadsheet with new data. Delays in PAHO Reports: We recognize there are delays in what PAHO reports and reports by media or local health organizations. However, in order to be fair and consistent in scoring submissions, the final scoring for accuracy and peak prediction will be judged against PAHO reports. The final scores for accuracy of forecast and peak prediction will be based on PAHO data, which will be considered ‘official’ four weeks after the challenge ends, in case there are revisions in PAHO reports. Solvers are encouraged to use all publicly available data sources including news articles and media reports, but should recognize that scoring will be based solely on what PAHO reports, and we cannot validate the accuracy of any information released to the media through non-PAHO channels. Solvers will not be scored for weeks where PAHO data is repeated due to a lag in reporting. However, solvers will be scored for instances where PAHO reports no cases (or the country is not present on the PAHO report for a particular week) and new cases arise during the period of evaluation. Total Cases Decrease between Weeks: In the case that PAHO data reports that the total number of cases decreases between weeks and the cause is not obvious i.e. error in data entry, those weeks will not be scored. Potential Errors in PAHO Reports: 1. Dominica Week 24 & 25: Reported EW 20 cases during Week 25, when Week 24 reported EW 22 data. Week 25 has been set to EW 22 data in the spreadsheet. 2. Sint Maarten Week 5: Reported EW 3 confirmed cases during Week 5. Week 5 has been set to EW 4 data in the spreadsheet. Negative Sign in Accuracy Scoring Formula: The negative sign has been removed in the formula for scoring accuracy. The threshold for eligible submissions remains at 500. This information will also be recorded in the Challenge Resources page. -Kevin InnoCentive Client Services Prithwish Chakraborty Graduate Research Assistant Dept. of Computer Science Discovery Analytics Center Virginia Polytechnic Institute and State University email: prithwi(a)vt.edu web: http://people.cs.vt.edu/~prithwi/

1 0

Re: [Darpa-chikun-challenge] CHIKV Challenge Submission #1
by M Elizabeth Halloran 01 Sep '14

01 Sep '14

Dear Prithwish, Thanks for sending the figures. They predict a lot of Chik cases coming up. This would make sense for a simple SIR model. What are the numbers after the legend where it says "Actual" ? I did not understand a couple of the graphs, for example Saint Martin and Saint Barthelemy. This looks like a great beginning. Best wishes, Betz On 8/31/14, 10:40 PM, Prithwish Chakraborty wrote: > Hi All, > > The predictions results have been pushed to the github repo. These > predictions are from a simple SIR model with no external data. > This baseline method is adapted from Elaine's code with the exception > being different optimizer (constrainedSLSQP > <http://www.pyopt.org/reference/optimizers.slsqp.html>) and using > different scaling around the peaks. > > You can find the figures at > https://github.com/prithwi/dac_chik/blob/master/data/results_sub1/fig/ > and the corresponding csv's in > https://github.com/prithwi/dac_chik/blob/master/data/results_sub1/csv/. > > There are predictions for 20 countries in there. Of the remaining, 30 > don't have sufficient data to fit SIR model. > (see > https://github.com/prithwi/dac_chik/blob/master/data/results_sub1/Notpredic… > ) while for the rest the SIR model gives very poor fit. > > Am attaching the plots herewith for your convenience (as I think many > of the people don't yet have access to the git repo). > I look forward to your comments on the predictions. > > Thanks > > P.S. > 1) Its a very simple model this time mainly due to shortage of time. > But hopefully with our experiences from this time we can add Weather, > Travel Data, etc in the next submission where we will have to predict > the peaks as well. > > 2) Codes have also been pushed to the repo > > Thanks > Prithwish Chakraborty > Graduate Research Assistant > Dept. of Computer Science > Discovery Analytics Center > Virginia Polytechnic Institute and State University > email: prithwi(a)vt.edu <mailto:prithwi@vt.edu> > web: http://people.cs.vt.edu/~prithwi/ > <http://people.cs.vt.edu/%7Eprithwi/> > > > _______________________________________________ > Darpa-chikun-challenge mailing list > Darpa-chikun-challenge(a)cs.vt.edu > https://mailman.cs.vt.edu/mailman/listinfo/darpa-chikun-challenge -- M. Elizabeth Halloran MD MPH DSc Professor Center Stat. & Quant. Infect. Dis. (CSQUID) Vaccine and Infectious Disease Institute Hutchinson Research Center and Department of Biostatistics University of Washington 1100 Fairview Ave N, M2-C200 Seattle, WA98109-1024 USA Phone: 206-667-2722 Fax: 206-667-4378

1 0

Re: [Darpa-chikun-challenge] CHIKV Challenge Submission #1
by Alessandro Vespignani 25 Aug '14

25 Aug '14

Hello Prithwish, No problem for the air travel dataset. Can we talk on monday (tomorrow) to define format, time scale etc.? Regards Alex Vespignani On Sat, Aug 23, 2014 at 2:50 AM, Prithwish Chakraborty <prithwi(a)vt.edu> wrote: > Hi All, > > As you may recall, the first submission is due on Sep 1 (in a weeks time). > For that, we have to send 6 months-ahead predictions for 55 > countries/regions (weekly case count). As was discussed during the first > meeting of the group, Elaine , James and I met today and here's what we > thought of doing. Please chime in with your thoughts. > > To summarize, we are going with country specific compartmental SEIR model > for baseline estimates. We can add inter-country transmission effects to the > same. However we would need airline data to do so (details in P.S.). Prof. > Vespignani, I was wondering whether your group can help in organizing this > dataset or perhaps this is already setup? > > For Textual sources we need some keyword list to filter for. Am attaching > the ones Elaine came up with very quickly (also added to wiki at > Data-Sources#keywords-for-textual-sources). It would be of great help, if > you all can add more. > > Thanks > Prithwish > > > <P.S> > Wiki/Repo Access > ============== > Just a reminder that the wiki for this challenge can be found at > https://github.com/prithwi/dac_chik/wiki. It has several pages already being > setup describing the project details (with link outs to folder in the repo > which contains the submission templates), models and data sources. If you > don't have access to it, please send me a mail with your github username so > that I can add you to the collaborators list. > > Baseline Model > ============= > 1. @Elaine will produce ODE based compartmental SEIR simulations to > generate forecasts. ETA Tuesday Aug 26. > > Salient points : > a) @Prithiwsh will provide population estimates. Baseline data already > setup. > b) Won't include transmission factors between countries. > > 2. @James and @Prithwish: Will try to get transmission factors between > countries based on the Nowcasting paper. > > Salient Points: **NEED AIRLINE TRAVEL DATA to create the transmission map** > > Data Sources > ========== > Data Sources | Assigned | Status > 1) Weather | James | may get ready before submission #1 > 2) Healthmap | Prithwish | ETA Monday Aug 25. > 3) Twitter | Prithwish | ETA Monday Aug 25. > 4) GST | Prithwish | ETA Tuesday Aug 26. > 5) Mosquito | Elaine | Has data for Dominican Republic. > 6) Air Travel | TBD | TBD > > Prithwish Chak > raborty > Graduate Resea > rch Assistant > Dept. of Compu > ter Science > Discovery Anal > ytics Center > Virginia Polyte > chnic Institute and State University > email: prithwi(a)vt.edu > web: http://people.cs.vt.edu/~prithwi/ > > _______________________________________________ > Darpa-chikun-challenge mailing list > Darpa-chikun-challenge(a)cs.vt.edu > https://mailman.cs.vt.edu/mailman/listinfo/darpa-chikun-challenge > -- --------------------------------------------- Sternberg Distinguished University Professor, Department of Physics College of Computer and Information Sciences Bouve' College of Health Sciences Northeastern University, Boston MA

2 2

Weather Bleed Started
by James Schlitt 23 Aug '14

23 Aug '14

Hi All, Hope all is going well. I just set up a weather scraper to pull some quick forecasts and historic data for us. For each country listed on the challenge, I am pulling data once every three days via forecast.io from Jan 1, 2014 to March 1, 2015 from the 2 most populated cities of the given country. This should take about 3 days to complete with cities being pulled in order of decreasing country population. If you need access to this data as it pulls, please send a gmail account and I'll share the file. The cities used as well as city metro area and country population data may be found here: https://docs.google.com/spreadsheets/d/1xvqgAF17Av7NuDLZ3PaZoaKV7fvdQ_VLfaY… Cheers, James Schlitt

1 0

Action items from today's meeting
by Naren Ramakrishnan 21 Aug '14

21 Aug '14

Folks, Great to get started on this challenge. Some action items: 1. Prithwish will setup a Github/Wiki page and share with all. 2. Prithwish and Elaine to coordinate to generate forecasts with a baseline model for the first prediction to be delivered. 3. Alex and Elaine/John - pl consult with your colleague(s) if they would like to join the team, and I will add them to the list. 4. John to share some recent relevant papers on CHIKV. More coming soon, Best, Naren -- Naren Ramakrishnan Thomas L. Phillips Professor of Engineering Department of Computer Science http://www.cs.vt.edu/~naren Director, Discovery Analytics Center, Virginia Tech http://dac.cs.vt.edu

3 4