基于免疫原理和Boosting机制的模糊分类规则挖掘算法

Algorithm for Mining Fuzzy Classification Rules Based on Immune Principles and Boosting Mechanism

基于免疫原理和Boosting机制,提出了一种模糊分类规则挖掘算法.该算法主要借鉴于自然免疫系统中的克隆选择原理,通过抗体种群的演化来优化模糊规则.模糊规则库通过增量的方式产生,算法每次运行得到一个规则.Boosting机制用于调整训练数据的权值,使得新生成规则集中于当前未被覆盖或误分类的数据实例.仿真实验表明,所提算法可根据规则的分类精度来调整训练数据的权值,促进了模糊规则之间的协作关系,避免了规则之间相互冲突,提高了系统的分类精度.

An algorithm is proposed for mining fuzzy classification rules by using natural immune principles and boosting mechanism. The proposed algorithm is mainly inspired by the clonal selection principle of biological immune systems, and the population of antibodies is evolved to optimize fuzzy classification rules. The fuzzy classification rule base is generated in an incremental way, in which one rule is obtained in each run of the proposed algorithm. When one rule is generated, boosting mechanism is used to change the weights of the training instances so as to mine new rules that are focused on currently uncovered or misclassified instances. The proposed algorithm can promote the cooperation and avoid the conflict among fuzzy rules by using boosting mechanism. Compared to other relevant algorithms the proposed algorithm has better predictive accuracy.