Ember Alert

Inspiration

Recently, California was devastated by a series of wildfires. As a result, this led to thousands of dollars in damages and dozens of fatalities. Our goal was to create a web app that allows people to be more informed of what to do when preparing for a wildfire, letting them know beforehand if they are at risk for wildfires.

What it does

Our web app takes in a city or a zip code that the user supplies and uses the weather information (temperature, wind speed, and humidity) of that location to determine if that location is at risk for a wildfire. If the app detects that the location is at risk for a wildfire, the website directs the user to a page that advises them on what actions to take to prepare for the wildfire and to evacuate, if necessary.

How we built it

We used a logistic regression on the three variables (temperature, wind speed, and humidity) to find a good way to classify if a location was at risk for wildfire (modeled by either a 0 or a 1, 0 indicating low risk, 1 indicating high risk.) We trained this prediction model against data from prior wildfires in the United States. We then used IBM cloud to integrate this prediction model into our Wix website.

Challenges we ran into

We found that incorporating all of our results onto a Wix website was a lot more difficult than we had anticipated. This primarily came from trying to pass data from frontend to backend and vice versa. In addition, we faced challenges trying to control the flow of the logic in the program, trying to navigate some of the more advanced features of JavaScript Promises.

Accomplishments that we're proud of

We managed to create a machine learning model that works properly on data it's never seen. We also learned how to transfer code from our local system to the Wix servers. We also learned how to transfer code from one language, going from mathematical models in Python to scripting in JavaScript.

What we learned

We learned how to put data onto a framework that makes it possible to integrate multiple different sources of code into one cohesive application. We were all exposed to different pieces of software for building new applications like IBM Cloud, Wix code, and machine learning libraries from Python.

What's next for Ember Alert

We want to train a more accurate model for our classification algorithm by collecting more relevant information about the environmental factors of a given region. For example, more specific data on brush density, eliminating non-independent input variables, and more data to work with in general.