基于蚁群算法的车辆状态疲劳特征优化研究

Optimization of Vehicle State Fatigue Feature Based on Ant Colony Algorithm

针对疲劳驾驶引起的道路交通安全问题,提出一种改进蚁群优化(IACO)算法对车辆状态疲劳特征进行优化。将特征选择问题转化为全连接的无向图,引入Fisher分数与最大信息系数来提高搜索效率、降低特征冗余度。从车辆状态参数中提取疲劳特征,利用IACO算法对疲劳特征进行优化,得到最优疲劳特征子集。实验结果表明,IACO算法的SVM分类准确率为85.6%、KNN分类准确率为83.2%,均高于其他常用特征优化算法的分类结果,说明IACO算法对疲劳特征的优化性能高于其他常用特征优化算法。

A modified ant colony optimization(IACO)algorithm is proposed to optimize the fatigue characteristics of vehicle states in response to road traffic safety issues caused by fatigue driving.The feature selection problem is transformed into a fully connected undirected graph and fisher scores and maximum information coefficients are introduced to improve search efficiency and reduce feature redundancy.Fatigue features are extracted from the vehicle state parameters when drivers are driving and the fatigue features are optimized using the IACO algorithm,and thus the optimal subset of fatigue features are obtained.The experimental result shows that the SVM classification accuracy of fatigue features optimized by the IACO algorithm is 85.6%,and the KNN classification accuracy is 83.2%,both of which are higher than the classification results of other commonly used feature optimization algorithms.This indicates that the optimization performance of the IACO algorithm on fatigue features is higher than that of other commonly used feature optimization algorithms.