基于粗糙集的增强学习型分类器

为了提高分类的精确度，提出一种基于粗糙集理论的增强学习型分类器。采用分割算法对训练数据集中连续的属性进行离散处理；利用粗糙集理论获取约简集，从中选择一个能提供最高分类精确度的约简。对于不同的测试数据，由于离散属性值的变化，相同的约简可能达不到最高的分类精确度。为克服此问题，改进了Q学习算法，使其全面系统地解决离散化和特征选择问题，因此不同的属性可以学习到最佳的分割值，使相应的约简产生最大分类精确度。实验结果表明．该分类器能达到98％的精确度．与其它分类器相比．表现出较好的性能。

Reinforcement Learning Classifier Based on Rough Set

In order to improve the accuracy of classification, a reinforcement learning classifier based on rough set theory is pro- posed. First, the continuous attributes in the training data set are discretized by using segmentation algorithm. Second, reducts are obtained by using file rough set theory and finally, one of the reducts providing the highest classification accuracy is chosen. But for the different test data, the same reducts may not reach the highest accuracy of classification because of the changes of dis- crete attributes value. To overcome this problem, Q - learning algorithm of the reinforcement learning is modified and it can com- prehensively and systematically solve the problem of the discretization and feature selection and make different attributes to learn to the best cut value so that the corresponding reducts can produce the maximum accuracy of classification. Experimental results verify that the classifier achieves the accuracy of 98% and exhibits excellent performance compared with other classifiers.