Clusters and Traveling Fisherman

This paper describes a solution for the FishAI competition organized by Norwegian Artificial Intelligence Research Consortium. The competition consisted of three tasks: forecasting the location with the most fish caught by date, creating the optimal fishing plan, and providing a visualization for the fishermen to utilize these results. We compared multiple approaches for the forecasting task. The most promising approach approach to creating the optimal fishing plan was to complete k-means clustering and subsequently solving the traveling salesman problem for the most viable forecast locations. Our visualization consisted of an interactive map with efficient fishing locations and routes as well as key performance indicators. Our results suggest that forecasting the single best fishing location is difficult but that efficient fishing plans can be formulated based on forecasts for larger regions.