blueberry

Toronto estimates that there are upwards of 15,000 lots that are vacant or abandoned in the city. Simultaneously, nearly every homeless shelter is at maximum capacity, due to increasing rates of homelessness, combined with the strain of the virus. We tackled this problem, with the aim of finding a solution that would allow for smart city planning for allocation of region-specific homeless resources.

We selected, researched and combined several different Toronto government databases to generate our comprehensive dataset. The Toronto databases store a vast amount of neighbourhood statistics, resource addresses and information along with providing a nice API framework for accessing it. Some examples include mental health centre locations, educational service locations, social housing info and crime rates. Taking these attributes and more, we developed a proximity and region based dataset listing over 2000 unique resources across the city. The fields include resource locations, with attributes based on the neighbourhood-specific information from the Toronto databases, as well as proximity information obtained through the Google Maps API. We used Google Maps address ratings as a metric of 'popularity' of a given resource. With this database, we expect that a Machine Learning could be used to predict the best resources for the homeless community within a specific neighbourhood from an address.

We spent a large portion of our 24 hours brainstorming and discussing the challenging requirements of this problem. We wanted a labelling metric that would represent the "busy-ness" of a given location, so that we could determine what resources were especially required in different areas. Unfortunately, this data is not readily available and we had to settle for the Google Maps rating, which was also sometimes lacking for specific locations.

An enormous amount of time was devoted to pre-processing and organizing the data. We're excited that we were able to create a powerful, yet organized dataset that stores a diversity of information. We learned about dealing with location data efficiently, and how to handle large datasets effectively.

We want to be able to use this database to push for increased resources for the homeless community. We hope that this database can be applied to make the decision process of development in cities more fluid. Specifically, we hope to train an ML model to quantify the "impact metric" of various resource centres (i.e. mental health based, homeless shelters, food banks, financial services) based on a given location. With this project, we are emphasizing that resource creation in cities can be streamlined, and doesn't need to be bogged down by logistics and indecisiveness. Indeed, our organized and robust database can be used to determine how to allocate the most impactful resources to homeless communities around the city.

Built With

• ckan
• collab
• google-maps
• python