Myself and many of my friends are very privileged to grow up with a roof on our head but many children aren't and unfortunately sets them up for failure. The perpetual cycle of experiencing homelessness induces a significant amount of trauma in children that hinders their future success. Current solutions to the problem are very reactionary, I wanted to provide a preventative model that intervenes with children before it is too late.
What it does
LOTUS answers three very critical questions:
1) Where is a child expected to experience homelessness next? 2) Identify children at risk in a school. 3) Quantify the savings of investing now versus later in a child.
How I built it
Leveraging previous experience with implementing K Nearest Neighbors Algorithm, Softmax and Linear Regression, I was able to fairly easily build a model that would predict where a child may be homeless next based on open-source data that is available from the City of Toronto. This was great, but I knew that if I wanted to provide a unique solution that was truly preventative, I needed to find a way to collect day where childern are every day. That is in schools. Thus, leveraging modern sentiment analysis algorithms via a Support Vector Machine, I was able to prototype a algorithm for Twitter sentiment analysis that can be the foundation for other platforms like Instagram, TikTok, Facebook etc. Through collecting teacher forms, analyzing social media and search history, we are able to build a model that graphs polarity vs subjectivity to characterize a childs action as negative, positive or neutral. With this information we can then put children on a radar and intervene with them as a preventative measure before it is too late. This information is all sent to the government where the allocation of resources can be redistributed in an appropriate manner.
Challenges I ran into
At the beginning of the day, I had no idea how to implement or work with sentiment analysis. I spent about 7-8 hours understanding how it works and how to implement it on different platforms. I ended up learning how to use it specifically for Twitter. I KNEW THAT AT HACKATHONS STAKEHOLDER VALIDATION IS SOMETHING THAT IS OFTEN OVERLOOKED. I WENT TO THE CONVENET HOMELESSESS SHELTER IN TORONTO TO UNDERSTAND THE PROBLEM FURTHER AND I HAD A LOT OF GREAT INSIGHTS FROM THIS EXPERIENCE.
Accomplishments that I'm proud of
I am very excited to present my project as I do believe that is a feasible platform to be incorporated into the Toronto District School Boards as well as the City of Toronto to genuinely help children. This model lays the foundation for future work in aiding children to come out of homelessness.
What I learned
I learned a lot about the homlessness community. After reading through coutnless papers, visiting homeless shelters and communicating with teachers, I recognize that this community is suffering however, there is a lot of opportunity to improve with modern technology in place.
What's next for LOTUS
1) TEST the platform in the city of Toronto 2) DATA USAGE AGREEMENT- figure out how to work with the city and other institutions easily 3) Summer data - figure out a way to collect data in the summers.