Overview TimelineProblem Details Submission Deliverables Divisions Guidelines Prizes Rules

Overview

RMDS Lab is one of the largest data science community with the headquarter located in Los Angeles. Our community has grown rapidly since its inception in 2009 and RMDS has held many data science related activities. RMDS proudly announces our very first Data Science Competition and would like to invite all talents to participate

RMDS Lab is one of the largest data science community with the headquarter located in Los Angeles. Our community has grown rapidly since its inception in 2009 and RMDS has held many data science related activities. RMDS proudly announces our very first Data Science Competition and would like to invite all talents to participate

The Problem

Improving hurricane intensity forecast is urgently needed to reduce economic losses associated with landfalling hurricanes. Contestants are asked to create a statistical model to predict tropical cyclone (TC) intensity change in 24 hours using advanced machine learning techniques. Teams will have access to two sets of TC data from 1998 to 2010. One is the predictors operationally used in a statistical TC intensity forecast model. The other is satellite observations with spatially and temporally varying features of TC structures. The combined use of two datasets is expected to yield better results than using one dataset only. The model performance will be judged by the skills of the model in predicting TC rapid intensification (define as TC maximum wind speed change greater than 20 kt in 24 hours) occurrence for all years (See Kaplan et al. 2015). This competition provides teams the opportunity to showcase their knowledge and network with company representatives, who will serve as members of the judging panel. 

Download research paper here:
Kaplan, J. et al., Evaluating environmental impacts on tropical cyclone rapid intensification predictability utilizing statistical models, Weather Forecast. 30 1374–96, (2015).

Download dataset here:
https://grmds.org/competition-data

Help with getting started:

• Get dataset FAQ here: dataset_guide.pdf
• Recorded training session with expert panelists: https://youtu.be/fbYOnHr7fzA?t=277
• Training powerpoint: Data_introduction.pptx

Details

Participants will work in teams of up to 4. The competition will start on Oct 23rd and all groups will have one month to complete their project. There will be a live training session for spatial data analysis and the dataset on Monday Nov 4th at 10 AM PST. As final deliverable, all groups will submit their project along with an analytics report no later than Nov 26th (deadline extended). RMDS Lab’s advisory committee consisting of experts from NASA JPL, IBM, LMU, will then evaluate and select the finalist. The finalist will have an opportunity to present their project at RMDS Annual Conference starting on Dec 6th. We will also hold an award ceremony for the winners on the afternoon of Dec 7th.

The analysis tools include but are not limited to: R, Python, Spark, etc. The submitted prediction models will be evaluated using a test dataset. The team with the best PSS score will be the winner of the competition. When the PSS score is tied, the team with the lower FAR score will win. See more details about evaluation methods in the dataset_guide.pdf

Submission Deliverables

• Create a zip file the following:
  • CSV of your results with five columns, with this naming convention: date (timestamp YYYYMMDDHH), clat, clon, predicted_dvmax, predicted_RI
  • Source code including the trained model
  • A README file explaining how to run your codes. If you use Java or C++, please also include the commands you use to compile your code (we should be able to compile, if necessary, and run your code and see the output files generated.
  • Research Report in PDF
• Naming convention:
  • CSV of your results should be saved as solutions_yourUserName.csv
  • Zip file should be saved as competition_username.zip
• Remember to set your submissions as private

Upload your submission here:
https://grmds.org/node/add/projects

Divisions

The competition will be split in two divisions: one for students and one for professionals. All who are interested and passionate are welcome to join.

Guidelines

Stage 1: Registration

Each team member will register on RMDS Lab webpage. Be sure to check the box stating your intent to register for the data science competition. We will send out a confirmation email to all candidates upon successful registration. Add info about your team in the Google Form. If any questions, please email: cici@rmdslab.com.

Stage 2: Team work and submission

You have about one month to work on the project. Submissions must include all the coding, a CSV file of your prediction, and one separate project report in PDF. Please zip all your files together and upload the deliverable to RMDS website. The deadline for submission is Nov 26th 11:59 PST. Place the names of all team members on the file. Submission by any individual group member will represent the whole team.

Stage 3: Evaluation and Final Presentation

Our experts committee from NASA JPL, IBM, LMU will evaluate all project deliverables and select the finalist teams by Dec 1st. The finalist teams will be invited to make their final presentation at RMDS Lab’s annual conference IM Data on Dec 7th. The evaluation metrics detailed may be disclosed in the future announcement.

Prizes

Additional Prizes: All winning teams will also receive trophies, certificates, and internship opportunity at RMDS Lab. All participants in this contest will receive an additional ticket to our conference.

*:For teams outside of Los Angeles, RMDS Lab will provide travel assistance to representatives of the winning team to join the annual conference and award ceremony at Pasadena Convention Center on Dec 6-7.

Rules

Eligibility: All participants should be over the age of 18 at the time of entry. Residents from a country designated by the United States Treasury’s Office of Foreign Assets Control are not eligible to receive any prize in the competition.

Submissions: Submission of your original formatted script, a CSV of your prediction result, and a PDF of project report will be needed. You may not directly calculate dvmax from a future time stamps.  You need to follow the instruction on your registration page and submit the files to RMDS Lab website. If your submission does not qualify our requirement, we will send out a notification email ASAP. Fail to resubmit within 24 hours of notification will be deemed to be abstention.

Intellectual Property: Data provided in this competition is an intellectual property of RMDS Lab and should not be spread out without specific permission. Participants should use the data only for the purpose of this competition. All should not transmit, duplicate, distribute, or publish the data to any other platforms.

Code: Sharing of codes between different teams are prohibited in the competition.

By submitting your script, you agree that RMDS Lab has the right to share your coding as an open source on our website.

Prize Winning: After the winner announcement, each winning team will receive the prize in the form of a check or wire transfer within 30 days. Prize winners are responsible for any taxes, fees or other liability resulting from their receipt of a prize.

Reservation of Rights: RMDS Lab reserves the right of final explanation.