Henry Joutsijoki and Martti Juhola
Benthic macroinvertebrate, Classification, Support vector machine, Machine learning, Pattern recognition
Automated identification and classification of benthic macroinvertebrates has got little attention. The research of benthic macroinvertebrates not only increases the knowledge about them, but also improves the methods of water quality monitoring.
In this paper our object is to investigate, how well Decision Acyclic Graph Support Vector Machines suit for the automated benthic macroinvertebrate identification. DAGSVM in this application is very little used, so the paper brings genuinely new information about the benthic macroinvertebrate identification. We performed experimental tests with six feature sets and every feature set was tested with seven kernel functions. Our experiments
indicate that DAGSVM suits very well to the automated identification of benthic macroinvertebrates. Especially identification with the smallest and largest feature set achieved very good results.