Comparison of Selected Methods of Multi-parameter Data Visualization for Efficiency in Evaluating Classification Possibilities of Various Coal Types
 
More details
Hide details
1
Faculty of Mining and Geoengineering, AGH University of Science and Technology
 
2
AGH University of Science and Technology
 
 
Publication date: 2015-06-16
 
 
Corresponding author
Tomasz NIEDOBA   

Faculty of Mining and Geoengineering, AGH University of Science and Technology, al. Mickiewicza 30, 30-347 Krakow, Poland
 
 
Physicochem. Probl. Miner. Process. 2015;51(2):769-784
 
KEYWORDS
TOPICS
ABSTRACT
Methods of multi-parameter data visualization through the transformation of multidimensional space into two-dimensional one allow to present multidimensional data on computer screen, thus making it possible to conduct a qualitative analysis of this data in the most natural way for human – by a sense of sight. In the paper a comparison was made to show the efficiency of selected seven methods of multidimensional visualization and, further, to analyze data describing various coal type samples (31, 34.2 and 35). Each of the methods was verified by checking how precisely a coal type can be classified when a given method is applied. For this purpose, a special criterion was designed to allow an evaluation of the results obtained by means of each of these methods. All the methods had already been described thoroughly by the authors in other papers. Detailed information included presentation of methods, elaborated algorithms, accepted parameters for best results as well the results. The framework for the comparison of the analyzed multi-parameter visualization methods includes: observational tunnels method (Niedoba and Jamroz, 2013; Jamroz and Niedoba, 2014), multidimensional scaling MDS (Jamroz, 2014b), principal component analysis PCA (Niedoba, 2014), relevance maps (Niedoba, in press), autoassociative neural networks (Jamroz, 2014c), Kohonen maps (Jamroz and Niedoba, in press) and parallel coordinates method (Niedoba and Jamroz, 2013).
eISSN:2084-4735
ISSN:1643-1049
Journals System - logo
Scroll to top