Overboard

See our presentation here:

https://docs.google.com/presentation/d/1L7yw4sZfrxLJQQyn08R6L46hUP3aLHLKWssihYC4R4I/edit?usp=sharing

Inspiration

We wanted to do something fun and creative. As we were thinking of ideas, Ayush randomly suggested pirates. At first, we didn't think it was possible, but we did more digging and realized we could do a Datathon project themed around Pirates and plundering the seas!

What it does

It can determine the probability of there being a shipwreck at a given coordinate. This takes into account historical trade routes, pirate activity, ocean conditions, and several other factors.

How we built it

We used 6 different data sources to build the dataset that we based our project upon. We first took the currently available data on shipwrecks from two different sources. We combined them and then used Open Meteo API to gather the ocean conditions for all the given coordinates. We then further extended the data by accounting for proximity to land, pirate activity, and historical trade routes. Finally, we used logistic regression to train our model to predict the probability of a shipwreck at a given coordinate.

Challenges we ran into

One of the hardest parts of this project was gathering and compiling our data. Since we were combining multiple data sets, it took some work to clean the data. The bulk of our time was spent on collecting the data - via API calls and Python packages. We found clever methods and workarounds to maintain as much relevant information as possible while still extracting data from the relevant sources at an acceptable pace.

Accomplishments that we're proud of

Finding sources to extract the relevant data we needed was our first major milestone. As we combined these sources and incorporated more information, seeing our final data set in one piece was also a major accomplishment during our project.

What we learned

We learned a lot of maritime terms, and we also found out the struggles of making a dataset from combining multiple different sources and adding your own information. We also learned where many sunken treasures may lie!

What's next for Overboard

One proposal is to create a web application so that users can see where wrecks are most likely to be found. Another would be to expand our project into identify other types of significant sea structures