Data preprocess and fuzzy rule establishment are included in the fuzzy learning algorithm. A set of training instances are collected from the environment. Our task here is to generate automatically reasonable membership functionsandappropriatedecisionrulesfromthesetraining data, so that they can represent important features of the data set. In order to avoid the disturbance of ineffective information,all the data should be preprocessed in advance