基于多目标优化的超盒粒计算分类算法

粒的数量和分类错误率是粒计算互相冲突的两个目标,同时最小化这两个目标是不可能的.针对此,构造了多目标优化问题,分别建立分类超盒粒数量和训练错误率两个目标,通过多目标演化算法对该多目标优化问题进行求解,从而产生一系列分类超盒粒集.随机产生初始种群,多目标演化算法通过利用演化操作和反复迭代的方法,得到供用户选取不同性能的解集.

The Hyperbox Granular Computing Classification Algorithm Based on Multi-objective Optimization

Granule number and classification error rate are two conflicting objectives in granular computing,it is impossible to minimize the two objectives simultaneously. The multi-objective optimization including the number of granule number and classification error was formed and solved by multi-objective evolutionary algorithm,and a series of multi-hyperbox granule sets were achieved. The multi-objective evolutionary algorithm obtained the different solution set by initialization of population,evolution operation and iteration method. Users can select the solution according to their requirements.