论文部分内容阅读
为了提高案例推理(CBR)分类器的性能,提出一种基于可信度阈值优化的CBR评价分类方法.首先,通过一种可降低时间复杂度的改进型可信度评价策略对案例重用得到的建议解的可信度进行计算;然后,通过遗传算法(GA)对可信度阈值进行迭代寻优;接着,根据得到的优化阈值将目标案例及其建议解划分为可信集或不可信集;最后,对不可信集按多数重用原则进行分类结论的调整,从而实现可信的CBR评价分类.对比实验表明,改进的可信度评价策略能有效提高分类性能,从而可提高CBR分类器的决策与学习能力.
In order to improve the performance of case-based reasoning (CBR) classifier, a CBR classification method based on credibility threshold optimization is proposed.Firstly, through an improved credibility evaluation method that can reduce time complexity, Then, the credibility threshold is iteratively optimized by genetic algorithm (GA). Then, the target case and its proposed solution are divided into trusted set or untrusted set Finally, we adjust the classification conclusion according to the principle of majority reuse by using untrusted set, so as to achieve credible CBR classification.Contrast experiments show that the improved credibility evaluation strategy can effectively improve the classification performance, which can improve the CBR classifier Decision-making and learning ability.