Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
1180539 | Chemometrics and Intelligent Laboratory Systems | 2007 | 7 Pages |
Although many methods are devoted to the design of Radial Basis Function Networks (RBFN), the lack of automatic approaches makes it difficult to generate suitable models in industrial applications. The object of this paper therefore proposes a deterministic method able to automatically select leaders or prototypes on which the RBFN design can be developed. This technique combines clustering and fuzzy C-means algorithms adapted to supervised contexts, and was tested successfully in a real application for Medicinal Chemistry, which was a data set regrouping 581 molecules active in the Central Nervous System. A comparison between the results obtained by this approach and by other standard initialization methods showed that our algorithm clearly improved the classification ability of RBFN.