Using Machine Learning to Analyze and Predict the Movement of Horseshoe Crabs in Long Island Sound

Dr. Samah Senbel, Assistant Professor of Computer Science at Sacred Heart University will be describing her work on developing machine learning models to predict animal movement patterns. Please join us (virtually) on Monday, September 27th at 1pm to learn more about this important area of study in ecology, conservation and wildlife management.

Register Now

Connection details will be provided after registration.

Abstract: Developing models to predict animal movement patterns is an important area of study in ecology, conservation and wildlife management. Models can be used to decipher patterns in mark-recapture data and machine learning can help to make predictions about future animal movement patterns. Project Limulus (PL), a community research program, has been tracking the movement of tagged individuals in the population of American horseshoe crabs (Limulus polyphemus) in Long Island Sound since 1997. During the spring spawning season, horseshoe crabs are captured by hand in spawning areas along the Connecticut (CT) shoreline, tagged and then released. Recaptured horseshoe crabs give valuable information about their behavior, if they exhibit site fidelity and movement patterns around the Sound. In this paper, we tested various models to find the best predictor for the movement of spawning horseshoe crabs to shorelines in the Sound based on the observed movement activity in previous years. The dataset consists of all the previous horseshoe crab movements: initial longitude and latitude, sex, initial date, and recapture longitude and latitude and recapture date. This dataset has 19,219 recapture records covering twenty years of activity. We experimented with three different models: Linear Regression, Decision Tree, and Random Forest Regression models. We used the data for 2018 as our test set and the data of all previous years as our training set. The Random Forest Regression model proved to be the best predictive model for animal movement and resulted in the smallest RMSE and MAE, as well as the smallest maximum error in prediction. The predicted horseshoe crab locations can be targeted in the next season for recapturing previously tagged horseshoe crabs, which provides valuable information about their movement patterns. It also concentrates the scientists’ effort and time to find the maximum number of horseshoe crabs.

This talk is supported by a grant from the Ramapo College Foundation.

Categories: Data Science, Lecture Series, MSDS