Abstract | ||
---|---|---|
Using the technique of multidimensional scaling the paper demonstrates a method of visualizing a configuration of classes as it is perceived by a classifier. The methodology serves to assist the analysis of multi-class classification problems, where the final result of averaged accuracy or averaged error is not sufficient. The approach may be used to control and tune different classifiers applied to the same data set or a single classifier applied to different data sets. The results of such analyses may then be used for identifying the combinations of classes that proved to be worst recognized. |
Year | DOI | Venue |
---|---|---|
2004 | 10.1007/978-3-540-39985-8_12 | INTELLIGENT INFORMATION PROCESSING AND WEB MINING |
Keywords | Field | DocType |
confusion matrix | Data mining,Data set,Confusion matrix,Multidimensional scaling,Visualization,Computer science,Classifier (linguistics) | Conference |
ISSN | Citations | PageRank |
1615-3871 | 1 | 0.37 |
References | Authors | |
7 | 1 |
Name | Order | Citations | PageRank |
---|---|---|---|
Robert Susmaga | 1 | 370 | 33.32 |