A method for extracting approximate rules from neural network
Hiroshi Narazaki, T. Shigaki, Toshihiko Watanabe · 2002
A knowledge acquisition method using a multilayered neural network (NN) is described. After training the NN for a data classification problem, our method extracts approximate classification rules from the NN; the method gives readability to the NN in which the information is represented in a distributed and unreadable manner. We show an application of our method to the standardization of operation conditions for a sintering process in an iron and steel making plant.>