Selecting Representative Prototypes for Prediction the Oxygen Activity in Electric Arc Furnace

Selecting a set of representative prototypes in prediction systems enable us to generate prototype based rules (P-Rules), which constitute a very powerful means of providing domain experts with knowledge about the data and the process depicted by the data. P-rules has already proved very useful in classification tasks. This paper investigates application of P-rules to regression problems. The problem of our concern is prediction of oxygen activity in an electric arc furnace during steel scrap melting. For that purpose we use a new algorithm for determining prototype positions, which is based on conditional clustering. Also a comparison between the new algorithm and the classical clustering-based methods for prototype extraction is described.