Ensemble based majority voting for point-to-point measurements of Gyrodactylus species identification

In the 21^st Century, a key challenge in both wild and cultured fish populations for control and management of disease is to securely and consistently perform pathogen identification. To provide automated accurate classification for the challenging Gyrodactylus species, we introduce an ensemble based majority voting approach for their classification. In this system, an ensemble classification approach is created that utilizes a combination of multiple feature sets and classifiers for Gyrodactylus species identification. The classifier base makes use of K-Nearest Neighbor (K-NN) and Linear Discriminant Analysis (LDA) approaches; with three different feature sets used for successful multi-species classification, considering 25 point-to-point data measurements, as well as smaller feature sets chosen using different feature selection techniques. The results show that our proposed ensemble based approach is accurate and robust, with ensemble based majority voting of classifiers and feature sets together found to be more effective than only combining feature sets.