Inexact knowledge discovery using Fish-Net algorithm

In almost all the disciplines of science and technology the knowledge applied to describing certain laws or regularities is not always precise. Thus the study of inexact learning becomes necessary and important. This paper presents inexact knowledge discovery using the Fish-Net algorithm which is an inexact field learning method for inducing forecasting rules from data. The experimental results show that this method is especially useful when the data are noisy, erroneous and with missing values. The algorithm can also be applied to the learning of fuzzy classification rules. It is interesting that the derived rules are capable of achieving higher prediction rates on new unseen cases compared with exact learning methods