马氏田口系统的量子行为二进制粒子群特征选择优化方法

针对标准二进制粒子群用于马氏田口系统的特征选择优化时，存在迭代速度慢，容易陷入局部最优解等不足，提出一种改进的基于量子行为二进制粒子群的马氏田口系统变量选择优化方法。首先，为了规避可能存在的复共线性特性对距离度量结果的影响，本研究采用Gram-Schmidt正交化法计算马氏距离值，对系统进行标准化处理，对各属性向量进行正交化后计算各类别的马氏距离集合，通过ROC曲线确定系统分类的最佳阈值点，定义误分类率概念和被选择变量占比最小作为变量筛选标准，构建多目标的混合规划模型。运用改进的量子行为粒子群算法求解优化组合，为适应二值化的变量优化问题，算法基于概率对粒子进行二进制编码，求取目标函数的适应值，并完成粒子群的优化迭代过程。采用优化的变量组合，构建精简的马氏田口系统，建立度量预测模型，完成精确判别的任务。最后，以胎心分娩力造影术测量的胎儿健康诊断为例，对标准二进制粒子群算法和二进制量子粒子群优化算法进行对比验证，实验结果表明，本文方法可以有效地提升粒子的迭代速度和寻优精度，优化后的马氏田口系统的预测准确率明显提高。

Feature Selection Optimization for Mahalanobis-Taguchi System with Binary Quantum Behavior Particle Swarm

When the standard binary particle swarm is used for the feature selection optimization of Mahalanobis-Taguchi system, the computational speed is slow and the selected features combination of Mahalanobis-Taguchi system is prone to falling into the local optimal solution. To address these problems, a feature selection optimization method of Mahalanobis-Taguchi system based on an improved binary quantum behavior particle swarm was proposed. Firstly, in order to avoid the influence of complex collinearity for the distance metric, the Gram-Schmidt orthogonalization method was used to calculate the Mahalanobis distance value. Through the ROC curve, the optimal threshold point for the system classification was determined. The misclassification rate and the selected variables rate were defined and a multi-objective optimization model was built. Then, an improved quantum behavior particle swarm optimization algorithm was presented to solve the optimization model, which performs binary coding on the particle based on probability. Through the optimized features combination, a new Mahalanobis-Taguchi prediction system was established. Finally, the fetal health diagnosis was carried out. The experimental results showed that the improved quantum behavior particle swarm optimization algorithm could effectively enhance the iterative speed, and the optimized Mahalanobis-Taguchi system had the better prediction accuracy.