IronHacks Live: Info session #3 - COVID-19 Data Science Challenge

Hacking for innovation to solve grand challenges
Dr. Sabine Brunswicker | Elizabeth Thompson | Matt Harris| Jia Lin Cheoh |
Satyam Mukherjee | other RCODI team members
July 27th, 2020
Research Center for
Open Digital Innovation (RCODI)
Covid-19 Data Science Challenges – Powered by IronHacks:
Info Session #3
Award Number: #1462044

Agenda
▪ Background: Who we are
▪ What is IronHacks - its past and its future!
▪ The COVID-19 Data Science Challenges 2020
▪ The COVID-19 Challenge: Summer 2020: Tasks, process, etc.
▪ Q&A
Agenda

Connected
everything
Analytics/AI
/ML
Cloud
Open platforms for
collective innovation
Social
media
Mobile
Background
Digital technologies afford more distributed and open forms of organizing the design
and use of new products/services
RCODI aims to empower individuals to use digital technologies to solve grand
challenges

Background
We are group of researchers and practitioners with interdisciplinary skills

Dr. Sabine Brunswicker
Director, RCODI
Elizabeth Thompson
PhD Student
Researcher
Jia Lin Cheoh
PhD Student Researcher
Background
Get to know the core group of people behind IronHacks 2020
Matt Harris
Platform
Developer
Marlen Promann
Satyam Mukherjee
RedHat Fellow
Shatayu
Undergraduate Student
Researcher

About IronHacks:
What is it - The past and its future

In the past, we thought innovation is about special people walled off
behind the R&D walls of large companies...

A challenge
A digital innovation
Digital platforms offer new ways to give hands to individuals to tackle
difficult problems

Source: https://flic.kr/p/Wd54U
Ironhacks is a platform that allows people around the world
to use data to solve problems

AFFORDABLE
HOUSING
Build a website with
interactive
visualizations that
helps new students in
finding safe and
affordable housing
near their university.
HEALTHY
LIVING
Use open data and
develop an app that
helps citizens in
finding cheap and
seasonally fresh
vegetables from
local markets.
About IronHacks
The IronHacks tasks focus on societal challenges that matter to all of us
COVID-19
Build a statistical
model using social
movement,
distancing and
infection data to
predict COVID-19
impact.

About IronHacks
The IronHacks supports a new hacking experience -
it is not a traditional hack
Traditional ‘Hackathons’ The Purdue IronHacks
36 hrs of high energy; less creativity
~ 1 month: 5x high-energy hacking + time for
creativity
Proprietary code: no sharing with
others
Open access: shared and can build on others’
code and data models
Mentorship during the 36 hrs;
Feedback on final solution
5 iterations: hackers get valuable technical,
market, and user feedback from data science
experts; constant guidance
Constant physical presence and
interaction
Virtual presence and communication;
hackers can work on problems when they wish

INNOVATE WITH UNIQUE DATA: Use unique, real-world and actual datasets to
create novel, cool, and performative models and visualizations that tackle real
challenges
OUTPERFORM & LEARN FROM OTHERS: IronHacks is a competition, but you also
engage in community guided by support from our team
HACK VIRTUALLY AT SCALE: IronHacks offers a novel cloud-based environment that
allows you to perform data analysis at scale
GAIN & FAME IN MANY WAYS: Learn, improve your score, gain reputation and win
a prize.
About IronHacks
Multiphase: hack, breathe, learn, iterate, and win...

About IronHacks
In the past we focused on web-app development and mash-up designs using datahe
Key technologies/programming languages:
HTML, CSS, Javascript, D3.js
Key skills/performance focus
1) Web programming
2) Interactive mash-up design (map, charts,
user input)
3) Visualization quality & novelty

The IronHacks process is iterative; there are 3 to 5+ phases; at each phase
your progress will be evaluated
About IronHacks
*Note: This is a sample timeline

Registration
Cloud-based editor Forum
The IronHacks platform has a range of unique features to facilitate participants
during the hacking process...
About IronHacks
Dashboard

About IronHacks
In the platform, participants can utilize a no setup coding environment...
File structure
& templates
Editor to create source code
Hosted app (page view)
Easy submission & github integration

The platform is designed for scalability and integrated process supporting
automated scoring and expert rating
Scalability
Integrated process (no set up
environment)
User experience
Real-time feedback
Supports randomized field experiments
and trace data collection
About IronHacks

About IronHacks
Since 2015, we have been growing; we ran hacks in US, Colombia, and China
4
2015:
Pilot
2016:
Pilot
2016 2016 2017
22 26 68
200
Numbers represent participants per hack
2017 2018
Bubble represents the number of participants at
the IronHacks competition, globally.
108 130
2019
54

Empowering Citizens
Existing processes for open data contests/hacking primarily focus on triggering
competition and neglect the power of digital technologies
DIGITAL PLATFORMS
In our research we design
and examine digital
platforms with the goal to
increase the productivity of
programmers.
REAL WORLD
IMPACT:
Supporting programmers
to design models and
visualizations that
support real-world
decisionmaking.
CURRENT HACKING:
...neglect the constituting
role of digital platforms
that encourage collective
learning and augment
human creativity.
Brunswicker, S., Jensen, B., Song, Z., & Majchrzak, A. (2018). Transparency as design choice of open data
contests. Journal of the Association for Information Science and Technology, 69(10), 1205–1222.
https://doi.org/10.1002/asi.24033

Career
opportunities
IronHacks
Certificate
Prize
money
Class
Credit
About IronHacks
In the past, we were able to offer financial and non-financial incentives
Global
recognition

In the past, we worked with corporate and government sponsors; this allowed
individuals to gain reputation
About IronHacks

About IronHacks
Our past editor supported building webapps in the cloud
File structure
& templates
Editor to create source code
Hosted app (page view)
Easy submission & github integration

In the hack in spring 2019, the students faced a housing challenge: “Find a safe and
affordable place to live in New York City (NYC)”
Imagine you are living in New York City (NYC) as a new student at NYU Stern School of Business and you
want help others that have no knowledge about the city to find an area to live that is safe, affordable
and close to the university.”
TASK: Develop a mashup that integrates, analyses and interactively displays open data in order to support new
students to evaluate and rank the 59 habitable districts in New York. Your mashup should allow the new students
to rank and compare the various districts using a diverse set of decision parameters. Your app should allow students to
find the top 10 districts in New York City based single parameters ranking and the top 3 districts based on average of
the three following parameters:
1. Safety
2. Distance to NYU Stern School of Business (Note: the distance must be from the center of the district to NYU Stern
School of Business).
3. Affordability (Note: affordability levels of a district are based on the maximum number of low income defined
by New York City).
“Find me a safe and affordable district to live near the NYU Stern School of
Business, New York.”
About IronHacks

About IronHacks
Today, the IronHacks platform offers a state of the art data science environment: Your
workspace will be a JupyterLab
Key technologies/programming languages:
Jupyter notebook, BigQuery, R, Python, SQL
Key skills/performance focus:
● Descriptive data analysis
● Data visualization for decision making
● Predictive modeling (Regressions,time
series, simulations)
JupyterLab 1.0: Jupyter’s Next-Generation Notebook Interface
● JupyterLab is a web-based interactive development
environment for Jupyter notebooks, code, and data
● JupyterLab support a wide range of workflows in data
science, scientific computing, and machine learning.
● JupyterLab is extensible and modular

About IronHacks
...so we are somewhat similar to Kaggle :-) ...

The COVID-19 Data Science Challenges
– Powered by IronHacks

The COVID-19 Hack - Powered by IronHacks
Predictions are essential during this COVID-19 pandemic; however, they might change
from week to week due to non-linear behavior

COVID-19 Challenges
Powered by IronHacks
Summer and Fall 2020
Sign-up today to predict the COVID-19
social and economic risks for citizens in
Indiana and other states of the US based
on actual data
The COVID-19 Hack – Powered by IronHacks

Registration opens
Tutorials and
Q&A start
Start of
competition
(min 60 ideally 100
committed
participants)
Submission 1 Submission 2 Submission 3 Submission 4
Score
release 1
Score
release 2
Score
release 3
Both hacks will follow an iterative process
Award ceremony

In our hack, you will use a range of BIG data sets to predict visiting patterns at point of
interests/core places in different counties
Key datasets (subset of eligible ones)
1. Core Places Patterns
2. SafeGraph Places
Patterns
3. Social Distancing
Metrics
4. Transaction Data
1. Average
housing Price
per
Neighborhood
(county-level, historical)
(neighborhood level, yearly)
(POI level)
1. COVID-cases
2. COVID-deaths
(zip-code level,
daily updated)
1. Unemployment claims
(zip-code level)
2. Skill-data
To make sure that you are not struggling with the data, we have
invested pro-processing and have created BigQuery datasets

In addition to providing you with access to pre-processed datasets, we will also
preinstall a large number of libraries
Libraries in R
1. BigRQuery
2. DBI
3. Tidyverse packages ( e.g. dplyr, tibble, stringr)
4. Lubridate
5. Data.table
6. Matrix
7. Zoo
8. Xts
9. Stats
10. Lmtest
11. Car
12. Sandwich
13. AER
14. DYNLM
15. ADLR
16. Forecast
17. Tidymodels
18. Lme4/lme- Linear and Non-linear mixed effects models
19. Caret
20. Ggvis
21. htmlwidgets
1. Pandas
2. Numpy
3. Scipy
4. Scikit-Learn
5. Statsmodels
6. Matplotlib
7. Seaborn
8. Plotly
9. Bokeh
10. Pydot
11. ...
Libraries in Python

Your task will be about predictions using different pre-processed datasets; the task will
be revealed at the start of the hack
Example...
Develop and execute a Python or R code in a Jupyter notebook that predicts COVID-19
impact using the datasets available via BigQuery (e.g. spending data, mobility data):
1. Prediction using actual data: Build a model to predict the visitor count/foot traffic for
the coming week using a statistical model of your choice.
2. Visualization: Show your predictions so that end-users can understand the historic
and the future COVID-19 risk!

You will submit your results in a csv file along with your Jupyter notebook and other
complementary information
.ipynb
.ipynb

To score/rank the submission, we plan to judge the submissions in four categories; we
combine objective machine-readable metrics with human expert judging
Usefulness of
visualization
• Presents
predictions of top
crowded places in
a useful way
Performance Evaluation
Accuracy of
prediction
• Accuracy of
model in
predicting (Mean
Square Error)
Quality of modeling/
reproducibility
• Includes model
description/logic
• Reproducibility of
results
• Quality of code

We will offer tutorials specific for the task after registration has opened; in addition, a
compilation of important online resources will help you to get ready for the hack
Jupyter Lab
BigQuery &
Jupyter
Notebooks
SQL
Essentials
Tbd: Dates
& times
And lots of additional curated online links..

Top 3 Solutions
Improvement
Spirit
We will reward you in different categories, depending on your and your peers
performance
Brase, G. L. (2009). How different types of participant payments alter task performance. Judgment and
Decision Making, 4(5), 419.
In total, we plan to have 100 participants complete the hack; we plan for a prize budget of ~ 3000 USD
0 to 25 percentile
25 to 50 percentile
50 to 75 percentile
75 to 100 percentile
Increasing
pay
Other Award
Categories

There are a range of benefits
For Citizens:
A high-quality Covid-19
model based on real data and
key insights about metrics
like social distancing can help
keep families informed on
the best actions for them to
take for themselves and how
doing so can protect others!
For COVID-19 Research:
New insights into nature of COVID-19
from a social and economic
perspective
For Policymakers:
Policymakers can utilize IronHacks
results to help make critical
decisions on budgets, health
policy and other critical decisions
during challenging times.

Here are some of our partners for the COVID-19: Data Science Challenges

What should you do next?

We are suggesting a task that focuses on prediction of visitor counts in Indiana
“When and where are the most crowded places in our counties?”
The scenario: Imagine you are part of a government task force to assess the COVID-19
risk emerging from “crowding” at public places. Using data analysis/data science skills
you want to help the government to understand and predict the visitor account on a
weekly basis to support real-time decision-making.
Your tasks:
Develop and execute a Python or R code in a Jupyter notebook that presents the results
in numeric values and visualizations related to the following three subtasks:
1. Historic description: Where are the places with the greatest decrease/increase in
visitor count since March 2020 relative to a baseline of 2019?
2. Prediction for following week: Build a model to predict the visitor count/foot traffic
for the coming week using a statistical model of your choice.
3. Visualization: Show your predictions in a spatial-temporal way so that end-users can
understand the historic and the future COVID-19 risk!
The COVID-19 Hack - Powered by IronHacks

About Me and RCODI
I founded RCODI in 2014 as a university-wide center; today, it brings together about 20
researchers and fellows from different disciplines and countries
Advisory Board
Center Director
Core Faculty
Affiliated Faculty
Purdue
Affiliated Faculty in the
US, Europe, and Asia
Members (donations
and data transfer)
Academic, research
members
Performance metrics
• Intellectual contributions to
scholarship of digital
innovation (Publications)
• Research grant applications
and research expenditures
• Successful graduated
students in the area of
open digital innovation
• Undergraduate research
support/fellows
• International relationships
with academia and industry
• Endowments
• Real world impact and
advocacy

Connected
everything
Analytics/
Big Data
Cloud
Open platforms
for collective
innovation
Social
media
Mobile
About Me and RCODI
The phenomenon of open digital innovation motivates our research team
Open digital innovation describes open and collective forms of organizing the design
and use of new products, enabled through digital platforms/technologies.

About Me and RCODI
Scientifically we span the boundary of social sciences (including economics), data
science (focus network science) and information systems

Dr. Sabine
Brunswicker Director,
RCODI
Elizabeth Thompson
PhD Student
Researcher
Jia Lin Cheoh
IronHacks Team
Some of RCODI’s team members are involved in IronHacks 2020
Matt Harris
Platform
Developer
Marlen Promann
Satyam Mukherjee
RedHat Fellow
Shatayu
Undergraduate
Student Researcher

IronHacks Live: Info session #3 - COVID-19 Data Science Challenge

More Related Content

What's hot (20)

Similar to IronHacks Live: Info session #3 - COVID-19 Data Science Challenge (20)

More from Purdue RCODI (19)

Recently uploaded (20)

IronHacks Live: Info session #3 - COVID-19 Data Science Challenge