Biometrical Letters Vol. 51(1), 2014, pp. 1-12
|
A VARIANT OF GRAVITATIONAL CLASSIFICATION
Tomasz Góorecki1, Maciej Luczak2 1Faculty of Mathematics and Computer Science, Adam Mickiewicz University, Umultowska 87, 61-614 Poznań, Poland, e-mail: tomasz.gorecki@amu.edu.pl 2Faculty of Civil Engineering, Environmental and Geodetic Sciences, Koszalin University of Technology, Śniadeckich 2, 75-453 Koszalin, Poland, e-mail: mluczak@wilsig.tu.koszalin.pl |
In this article there is proposed a new two-parametrical variant of the gravitational classification method. We use the general idea of objects' behavior in a gravity field. Classification depends on a test object's motion in a gravity field of training points. To solve this motion problem, we use a simulation method. This classifier is compared to the 1NN method, because our method tends towards it for some parameter values. Experimental results on different data sets demonstrate an improvement in efficiency and that this approach outperforms the 1NN method by providing a significant reduction in the mean classification error rate.
machine learning, nearest neighbor method, dynamic classifier, gravitational classification
