Nonlinear approach in classification visualization and evaluation
Veslava Osińska, Piotr Bała · Nicolaus Copernicus University Repository (Nicolaus Copernicus University) · 2009
In this paper we have proposed the novel methodology to visualize classification scheme in informatics domain. We have mapped a documents collection of ACM (Association for Computing Machinery) Digital Library to a sphere surface. Two main stages of visualization processes complement one another: classification and clusterization. Primarily classified documents were visualized and their further clusterization by means of keywords was crucial in evaluation process. For clusters analysis of given visualization maps nonlinear digital filtering techniques were applied. The clusters of keywords were characterized by a local accuracy. Obtained semantic map was included to validation process.