论文部分内容阅读
为提高属性约简算法处理含噪音和不确定大数据的性能,提出了一种基于协同进化云的属性集成多代理约简算法(CCAEMR).该算法首先基于MapReduce机制设计协同进化云框架,将整个种群分解成多个具有自适应规模的协同进化子种群,通过子种群的共享奖酬来加速属性约简实现.然后,构造了一种协同精英优化的多代理集成策略,确保划分的子种群能够充分探索交叠属性子集之间的相关性和相互依赖性,且具有较强的抗噪音性能,这些代理能保持在稳定的精英地区且取得了最佳收益.实验结果表明:所提出的CCAEMR算法在解决大规模和不确定复杂噪音数据的属性约简时具有更好的效率和适用性.
In order to improve the performance of attribute reduction algorithm to deal with noisy and uncertain big data, a attribute-integrated multi-agent reduction algorithm (CCAEMR) based on co-evolution cloud is proposed. The algorithm first designs a co-evolution cloud framework based on MapReduce mechanism, The whole population is decomposed into a number of co-evolutionary sub-populations with adaptive scale, and the attribute reduction is accelerated through the sharing reward of sub-populations.Then, a collaborative elite-optimized multi-agent integration strategy is constructed to ensure that the sub-population Can fully explore the correlation and interdependence between overlapping sub-sets of attributes, and has strong anti-noise performance, these agents can maintain a stable and elite regions and achieved the best benefit.The experimental results show that: the proposed The CCAEMR algorithm has better efficiency and applicability in solving the attribute reduction of large-scale and uncertain complex noise data.