FishNet

Inspiration

Over 90% of the United States' lobster supply comes from Maine 60% of the United States' scallop supply comes from Massachusetts 50% of the northeastern region squid comes from Rhode Island

Localized weather volatility in New England materially impacts national seafood pricing in the US and we want seafood distributers to make the most money they can so they can support local fishing initiatives giving them the best purchase price.

What it does

View historic fish economics data and forecast the future price of Lobster, Scallops, and Squid for Rhode Island, Massachusetts, and Maine. You can also ask questions regarding seafood sales or the effects of certain weather on pricing. We also added price-volume elasticity and price volatility so that users can determine if a price change has to do with a certain season, weather condition, or other economic factors.

How we built it

We took NOAA Datasets from the past 5 years in Rhode Island, Massachusetts, and Maine combined with a fish price dataset to create a a historic chart of seafood pricing for Lobster, Scallops, and Squid. We used Claude Code for helping create the interface and combine our dataset together. We are Using MongoDB to store our historic data and using OpenWeather to get the next two weeks of weather. Then we use Gemini's chatbot feature so that users can ask questions about the dataset or about short-term predictive pricing.

Challenges we ran into

Some challenges we ran into is finding datasets that had the same time ranges as well as frequency, some datasets we were searching had daily information while others had weekly or monthly. To solve this we combined the the datasets and used Claude to extrapolate the weekly data from over 50 years worth of fish prices into daily and introducing noise to make it more realistic.

Accomplishments that we're proud of

We have trained an XGB Regressor model on the historic data that we can then take the two-week weather forecast to determine what kind of effect it may have on a specific species seafood price. We also enjoyed experimenting with MongoDB rather than using a service like Firebase.

What we learned

We learned that Claude can help us engineer solutions we couldn't think were possible in a short period of time, after taking the workshop with Arnell he taught us how to prompt properly and use it to be efficient with our time by having the AI act as different senior roles to get better insights and plan out our architecture. Then recursively feeding its input back into itself to develop ideas from different perspectives.

What's next for FishNet

With funding we plan on getting access to private ocean buoy information to get higher fidelity data along with paying OpenWeather to get longer and more detailed weather forecasting data in order to make our model be able to predict for a further period of time. If we get funding we could go to the Seafood Expo of North America in Boston to promote our product nationwide distributors to get their feedback, as well as doing direct outreach to dockside buyers so we can understand how they get their pricing.