基于Boosting的模糊分类规则集成学习及应用

由于难于获得先验知识,样本可分性差,辐射源识别很难达到很高的识别率。结合AdaBoost算法和遗传算法,提出了一种模糊分类规则的迭代学习方法。在每轮迭代训练过程中,算法通过调整训练样本的分布,利用遗传算法产生分类规则。减少分类规则能够正确分类样本的权值,使得新产生的分类规则重点考虑难于分类和拒识的样本。在规则学习的适应度函数中考虑训练实例的分布,使模糊分类规则在产生阶段就考虑相互之间的协作,改善了模糊分类规则的整体识别能力。辐射源识别实验结果表明,该方法具有良好的分类识别性能。

Ensemble Learning and Application of Fuzzy Classification Rules Based on Boosting

Because available knowledge is hard to obtain and the separability of instances is bad, the classification of Radiant Point has a low recognition rate. An iterative learning method of fuzzy classification rules is presented based on the combination of AdaBoost algorithm and Genetic algorithm. At each iteration training of AdaBoost algorithm, the distribution of training instances are adjusted on which classification rules are created by Genetic algorithm. The weights of the training instances that are classified correctly by available rules are reduced, so that the new fuzzy rule focuses on the misestimate or uncovered instances. Because the distribution of training instances are attached to computing of the fitness function and the collaboration of rules is taken into account during producing rules. The classification performance of the multiple classifiers ensemble based on the fuzzy rules is improved. In Radiant Point experiments, this algorithm shows good recognition rate.