Machine learning for species classification of the invasive Centaurea Jacea hybrid complex

Presenter's Name(s)

Sophie Linde

Conference Year

2023

Abstract

Effectively managing biodiversity requires accurate taxonomic classification, which is challenging when species look similarly. Centaurea jacea and C. nigra are invasive weeds that hybridize extensively, leading to phenotypic variability that blurs easy classification. Using measurements and imagery from individual flowering heads and associated genetic ancestry estimates, I assessed the utility of machine learning models (decision trees, support vector machines, random forests, and neural networks) for species classification. Models were most accurate at separating C. nigra from C. jacea and hybrids. These models are not currently reliable for classifying Centaurea, but if refined, could become a helpful tool for their management.

Primary Faculty Mentor Name

Stephen Keller

Status

Undergraduate

Student College

College of Arts and Sciences

Second Student College

Patrick Leahy Honors College

Program/Major

Biological Science

Primary Research Category

Life Sciences

Abstract only.

Share

COinS
 

Machine learning for species classification of the invasive Centaurea Jacea hybrid complex

Effectively managing biodiversity requires accurate taxonomic classification, which is challenging when species look similarly. Centaurea jacea and C. nigra are invasive weeds that hybridize extensively, leading to phenotypic variability that blurs easy classification. Using measurements and imagery from individual flowering heads and associated genetic ancestry estimates, I assessed the utility of machine learning models (decision trees, support vector machines, random forests, and neural networks) for species classification. Models were most accurate at separating C. nigra from C. jacea and hybrids. These models are not currently reliable for classifying Centaurea, but if refined, could become a helpful tool for their management.