基于微粒群的案例推理方法研究

Research of CBR based on particle swarm optimization

距离测度是案例检索的关键问题,它直接影响案例检索精度.针对距离测度进行研究,提出一种基于微粒群方法的自学习距离测度,并将该自学习距离测度引入案例推理中,使案例推理在处理由相关属性表述的案例时有了合理的解决方法,从而扩展了案例推理的应用范围.最后,利用实际数据与UCI数据对基于新距离测度的案例推理技术进行了仿真实验,实验结果表明,与其他方法相比,该方法可以提高案例检索的准确性.

Distance measure is the key issue in case-based reasoning（CBR）,which influences the accuracy of case retrieval directly.For distance measure,a learning distance measure based on particle swarm optimization is proposed.The application range of CBR is extended by introducing leaning distance measure into CBR technology for the first time,which makes CBR technology have reasonable method to deal with the cases with correlative attributes.Finally,the simulation is conducted with real data and UCI data.The result shows that,compared with the other methods,this distance measure improves the accuracy of case retrieval.