-
-
The homepage for our web app
-
A sample prediction, showing both algorithm and AI model confidence
-
A summary of our project structure
-
Profile Page
Inspiration
It’s been known for years that impoverished neighborhoods experience much higher rates of chronic diseases, such asthma, diabetes, and more. Some of these conditions are adversely affected by climate patterns, like Air Quality Index and temperature. We decided to make a tool to assist these communities in this regard.
What it does
Our web app provides a user friendly portfolio page, after they register and log in. The user then provides their location and health conditions, which will be stored into a database along with other information. They will then have access to up to date risk analysis, through email updates or directly on our app, based on their health conditions and climate data, with 3 different levels of warning. Finally, the user can provide feedback on whether the recommendation was correct, improving our systems performance.
How I built it
Our app was built through the Django framework, which maintains the security of our product. We integrated a SQL database to keep track of patient information, including health conditions, over time. We also added an email server for more convenience. We use the Meteomatics and Weatherbit API’s to access climate data. Using all this data, our app runs a non-heuristic algorithm and a machine learning model (made from tensorflow and keras) to predict risk. The model can be improved through feedback from the user. Finally, we host our app on an Apache server.
Accomplishments that I'm proud of
We are proud of integrating our SQL database, which makes access to location and health data more efficient. We are also proud of our model’s ability to improve over time from user feedback. A model is only as good as the data it is given. Therefore, the ability to continuously retrain on new data drastically increases the potential for our model.
What I learned
We researched the effects of weather on numerous chronic diseases in order to create an algorithm that effectively takes into consideration. We also learned a lot about how to use Django to make and deploy our web app. We had some trouble with file paths when we migrated our app into the server, but we solved this in time.
What's next for WeatherOrNot
We will continuously train our machine learning model until it eventually replaces our algorithm. Currently, we only have about 20 diseases we can analyze; we plan to increase the size of this list. We also plan to use more climate data to better evaluate the risk present at any given day.
Log in or sign up for Devpost to join the conversation.