Ecological state evaluation of lake ecosystems revisited: Latent variables with kSVM algorithm approach for assessment automatization and data comprehension
Automated and reproducible methodology for assessing the ecological condition of lakes is essential for effective monitoring and facilitating the decision-making process aimed at achieving the stated environmental goals. At the same time, multidimensional measurement datasets are often an obstacle to drawing insightful conclusions, thus becoming an incentive for overly simplified analyzes. In this article, a set of measurements and ecological status assessment results for a collection of 499
... ollection of 499 lakes in Poland was used. Expert assessment process was recreated using the supervised kernel Support Vector Machine algorithm on dataset with reduced dimensionality, thus a model that automates the ecological assessment process was obtained. The use of the explanatory skill of latent variables made it possible to present the assessed objects along with their position in individual classes. The visualization of the results in reduced dimensionality increased, without interfering with the size of the classes, the informative evaluation potential, which should be considered as an acompanying assessment parameter in the future. The primary target of this paper is the ecological expert coping with automatization of assessment process and obtaining latent information for sense-making visual comprehension during consultations regarding ecosystem-oriented ecological decision making.