Visual concept detection by stacked generalization

Tekin, Mashar
In this thesis, we propose a new Stacked Generalization method, called Fuzzy Stacked Generalized Ranking Optimizer, to optimize the ranking performances of visual concept detection systems. In the proposed method, fuzzy k-NN classifiers are employed in the base-layer. Then, a classifier selection algorithm is employed to select the classifiers which will be combined in meta-layer. Finally, the results of the selected classifiers are combined and classified by a fuzzy k-NN meta classifier. In the experiments, the proposed method performs better than the state of the art ensemble learning methods.
Citation Formats
M. Tekin, “Visual concept detection by stacked generalization,” M.S. - Master of Science, 2014.