Prototype Extraction of a Single-Class Area for the Condensed 1-NN Rule
The paper presents a new condensation algorithm based on the idea of a sample representativeness. For each sample in a dataset a representative measure is counted. Starting with samples with the highest value of the measure, each sample and all its voters (which constitute single-class area) are condensed in one averaged prototype-sample. The algorithm is tested on nine well-known datasets and compared with Jozwik’s condensation methods.
Unable to display preview. Download preview PDF.
- 9.The ELENA Project Real Databases, http://www.dice.ucl.ac.be/neural-nets/Research/Projects/ELENA/databases/REAL/
- 10.Theodoridis, S., Koutroumbas, K.: Pattern Recognition, 3rd edn. Academic Press - Elsevier, USA (2006)Google Scholar