Training workflow of machine learning model
Testing workflow of machine learning model
Every individual in need deserves a place to sleep and path to avoid, or at the very least, move out of homelessness to financial stability.
What it does
Our system provides a way for managers of homeless shelter to add, track and update number of beds at a particular shelter location with a click of a button. Homeless people can go to our webpage and see the number of available beds at a particular shelter. Additionally, the second part of our platform helps automate the hiring process and continues to consider an individual for new jobs as their own skills increase. Finally, we would feed the data into a machine learning framework that would run analytics on locations where there are many homeless for the purpose of equipping the city with more data for preventative actions. The machine learning framework could also be used to pull data on those coming into the shelters to help better understand what an individual who may be at high risk of homelessness looks like before they even get to a shelter. The hiring data would also be fed into the machine learning framework to help drive better hiring decisions with time.
How we built it
The main component of our system addresses the issue of providing Emergency Shelter and Electronic Referrals/Data Sharing. For this component, we used Twitter to broadcast and store data, removing the need for server administration and easing administrative overhead. Data is then displayed on a Mapbox map using Leaflet. The local machines at the Emergency Shelter would have a clicker installed - someone hits + when a bed opens, a - when a bed is taken and that pushes to twitter. For the job hiring platform, we have started mapping the skills required for the types of roles those coming through the job training at St. Patrick's Center that can be updated as part of the intake process for those entering the 16 week job training program. This hiring platform would continue to consider an individual for new jobs as their own skill base grows. For the initial prototype, we are using basic web forms to fill out skill set that get exported to CSV and checked against the CSV that the hiring companies job checklist would generate. The machine learning framework is based in Azure and we used the mock data this weekend, but would pull the data from the digital forms that caseworkers already fill out.
Challenges we ran into
Working around API rate limits. Poor API/library documentation.
Accomplishments that we're proud of
We were able to address all four main issues faced by the homeless. The fact that this platform could help people and shelters find where there are available beds and that the second component of the platform automates the job hiring process and promotes upward mobility is very exciting. We are passionate about the idea of helping people escape poverty, not just find resources, which is why we married the two. Incorporating machine learning helps drive data analysis for prevention and better hiring to a new level.