DisasterFreeSpork

Inspiration

We wanted to help people during and before natural disasters so they could plan for any problems. We heard about people getting stranded and being confused about evacuation plans in Houston and wanted to develop an application that could solve it.

What it does

We have a huge database of natural disasters that have occurred in the United States since 1953. Based on this data we can determine if a natural disaster is likely to occur in a certain place and month. This was done with a BinaryLogisticalRegression machine learning model. Using this data we can also predict the most likely type of natural disaster that may occur.

When we know that a natural disaster may occur, we notify the user and then map them to the nearest shelter based on their GPS coordinates. This data is saved offline so in the event that cell towers are down, the user will have a route to a safe place.

How we built it

This was built using around 5 different languages and by all 4 members of our team. Josh and Joe focused on the web app while Ryan and Adam did the data analysis. We tried many types of classifiers and with the help of a mentor, agreed upon the BinaryLogisticalRegression model because it categorizes each event as a 0 or 1 which makes classifying much more accurate and efficient.

Josh created the web application in HTML, Javascript, and CSS which handled the mapping using the Google Maps API. He also did some logic on determining which natural disaster was most likely based on outputs from the Regressor. Joe worked on data transformation which allows the python data to transform into data parseable by the web app using java.

Challenges I ran into

Main challenge we ran into was figuring out the Regressor because at first none of the results were correct. With the help of some mentors we got this sorted out. Another challenge we ran into was figuring out how to use a ML regressor because none of us had super extensive knowledge into the topic. Formatting data was also a very challenging that took teamwork and lots of debugging to solve.

Accomplishments that we're proud of

We are proud of our web application and it's current functionality and also the cool Machine Learning algorithm we made.

What we learned

We learned the basics of Machine Learning and gained experience with processing large amounts of data that correspond to a few outputs.

What's next for DisasterFreeSpork

Improve upon the data collection methods and make the pattern recognition more robust. Add more fail safes to code and system so that disasters will be easier to navigate.

Built With

• css3
• html5
• java
• javascript
• python
• sklearn
• webpack