Inspiration
Climate change is a serious problem facing our society, especially in larger cities like Los Angeles and San Diego. However, one of the largest solutions that can help combat climate changes that city governments can do is building out valuable green spaces that allow for plants to thrive and filter the atmosphere and provide environment experiences for citizens.
What is does
The app reads from a machine learning Python model to predict whether cities would benefit from greenspaces given air quality, temperature, population, etc. From there, the user can observe the city using a Google Maps preview and make further decisions about the usefulness of greenspace from insights and recommendations from GPT-3.5-Turbo.
How we built it
We collected several attributes about San Diego cities that would determine whether greenspaces would be benefical for that community, including temperature, air quality and population, and labeled each city based on whether they have 3 or more parks. We then used this data to train an machine learning classification model to classify Los Angeles cities with characteristics that would need 3 or more parks. This is all presented in a Node.js application built with the EJS templating tool and Bootstrap CSS framework, with OpenAI GPT capabilities and Google Maps previews to help the user visualize the information.
Challenges we ran into
Initially, we had a problem collecting a large amount of training data efficiently and quickly, however we were able to solve this by pre-collecting data and websites with data that we wanted and feeding a chatbot to generate large CSVs to train off of, which resulted in being accurate for our purposes. Additionally, we ran into problems dealing with the OpenAI API, however this was caused through problems with permissions and were eventually resolved.
Accomplishments that we're proud of
Most importantly, we're proud of our machine learning model that has consistently held an accuracy of
What we learned
As a team, we all learned more about building machine learning models, as this was most of our's first time working with Pandas and Numpy. Additionally, we all learned much more about the impacts of climate change and how plants and green spaces can help minimize that effect.
What's next for GreenSpotter
If we were to continue development on GreenSpotter, the most important part would be collecting more organized training information from both San Diego and Los Angeles to train our models. Additionally, it would be useful to better fit the ML models to have a higher accuracy and allow for more insights in the user interface.
Log in or sign up for Devpost to join the conversation.