Prediction method with the variable threshold based on fuzzy association rules

Quantitative attributes are partitioned into several fuzzy sets by fuzzy c-means algorithm, and the search technology of a priori algorithm is improved to discover interesting fuzzy association rules. Then, the prediction method with the variable threshold based on the fuzzy association rules is presented. In this prediction method, a little error between prediction value and actual value is allowed. When the error is less than a given threshold, prediction value is regarded as acceptable or rational. The parameters of triangular fuzzy numbers are adjusted to improve the prediction accuracy by the genetic algorithm last. This prediction method can obtain the different prediction accuracy corresponding to the different error threshold chosen by the users, so it is more flexible and effective.