A Two-step Case-based Reasoning Method Based on Attributes Reduction for Predicting the Endpoint Phosphorus Content

Abstract

Case-Based Reasoning (CBR) system is a kind of solving paradigm based on the past successful cases to get the solution for the current problem. When CBR is applied in complex industrial processes, solving efficiency is often not high due to too many influence factors involved. So it is necessary to reduce the number of attributes involved in CBR system for the fast modern industrial production, such as steel-making and continues casting process. A two-step CBR method is proposed for predicting the endpoint phosphorus content in BOF efficiently. First, the genetic algorithm is applied to find the optimal attributes subset based on the evaluation method of Correlation-based Feature Selection (CFS). Then, CBR system is applied for solving this problem with the reduced attributes. There are two kinds of similarity calculation method based on the euclidean distance and the gray distance, and two kinds of the weight decision method based on the even weight and the entropy weight for this CBR system. Four groups of experiment results show that the two-step CBR method has much more efficiency than the single CBR method, while maintaining almost the same prediction precision. The two-step CBR method can be used in the fast industrial process more efficiently.