基于量子云模型反馈的协同精英属性均衡优势集成约简

为进一步增强进化算法在最小属性约简过程中全局求解性能,提出了一种基于量子云模型反馈的协同精英属性均衡优势集成约简算法,该算法首先设计一种基于云模型反馈的量子自适应旋转角调整策略,使量子蛙群精英在云模型定性知识和罚因子反馈指导下自适应控制属性搜索空间范围;然后构建一种有限理性区域下协同精英均衡优势属性分解框架,在动态精英演化区域内使参与属性约简的量子蛙群精英在平均权重裕度下协同化达到Nash均衡优势区域;最后量子蛙群精英采用集成化操作机制在各自均衡优势区域内协同提取属性约简子集,从而稳定取得全局最优约简特征集。实验结果表明本文算法求解全局最优属性约简集效率、精度和稳定性等具有明显优势,应用到孕龄新生儿脑MRIs电子病历分割进一步表明该算法具有较强的应用性能。

Quantum-cloud-feedback-based attribute equilibrium dominance ensemble reduction with co-evolutionary elitists

In order to further enhance the global performance of evolutionary algorithms for solving minimum attribute reduction, a quantum-cloud-feedback-based attribute equilibrium dominance ensemble reduction algorithm (QCAEDER) with co-evolutionary elitists is proposed. First, an adaptive strategy of quantum revolving angle update operation based on cloud mode feedback is designed, so that the search space scope of quantum frog elitists can be adaptively controlled under the guidance of qualitative knowledge of cloud model and penalty factor feedback. Second, the attribute decomposition framework of co-evolutionary elitists with equilibrium dominance under the bounded rationality regions is constructed, in order to assist the quantum frog elitists necessary to attain the stable status of Nash equilibrium dominance by the average weighted credits. Third, the quantum frog elitists can extract attribute reduction subsets in the respective regions of equilibrium dominance by using the ensemble operation mechanism. Consequently, the global optimal solution of ensemble feature set can be achieved stably. The experimental results show the proposed algorithm has achieved a higher performance on the efficiency, precision and stability of global optimal attribute reduction. Furthermore, the validation performed on brain MRIs electronic medical records of gestational age neonate has demonstrates its strong advantage for real-world applications.