论文部分内容阅读
针对非平衡数据集分类中“少数类样本精度难以提高”这一瓶颈问题,提出了一种基于协同进化机制的欠采样方法.此方法将少数类样本与多数类样本划分为两类种群,采用种群协同进化原理,利用提出的动态交叉变异算子自适应协同进化过程,实现种群间自动调节和自动适应.仿真试验结果表明,此采样方法增强了局部随机搜索能力,改善了种群的分布特性,加强了算法的全局收敛能力,在不降低多数类样本分类性能的基础上有效提高了少数类样本的精度.与其他经典重采样方法相比,本文办法抗噪能力好,具有更强的鲁棒性.
Aiming at the bottleneck problem of “the precision of minority samples is not improved” in unbalanced data set classification, an under sampling method based on co-evolutionary mechanism is proposed, in which the minority samples and majority samples are divided into two groups , Using the principle of population co-evolution, using the proposed dynamic cross-mutation operator adaptive co-evolutionary process to achieve automatic adjustment and automatic adaptation among populations.The simulation results show that this sampling method enhances the local random search ability and improves the population distribution Which enhances the global convergence ability of the algorithm and effectively improves the precision of a few types of samples without reducing the classification performance of most types of samples.Compared with other classical resampling methods, this method has better anti-noise ability and stronger Robustness.