基于最小二乘支持向量回归建模方法的人机系统操作员功能状态分析

目的建立具有很强预测能力的数学模型来准确评估人机系统操作员功能状态(Operator Function-al States,OFS)。方法基于采集到的一系列操作员电生理信号及性能数据,采用最小二乘支持向量机(Least Squares Support Vector Machine,LSSVM)方法对OFS建模。通过网格搜索和10-折交叉验证方法对模型参数进行优化,并将LSSVM与基于遗传算法的模糊建模方法进行比较。结果模型基本能反映OFS的实际变化趋势,输出误差在可接受的范围之内且与基于遗传算法的模糊建模方法得到的模型输出误差相比较小。结论 LSSVM方法具有更好的泛化性能,将其用于OFS评估是有效的。

Analysis on Operator Functional State of Human-Machine System Based on Approach of LSSVM Regressive Model

Objective To construct an optimum mathematical model to estimate Operator Functional State(OFS) in a human-machine system.Methods This paper adopted Least Squares Support Vector Machine(LSSVM) approach to construct OFS models with their multiple physiological and performance data.The model parameters were optimized with grid-search and 10-fold cross validation techniques.The modeling results of the LSSVM approach was compared with those of Genetic-Algorithms-based Mamdani(GA-Mamdani)-type fuzzy modeling method.Results The LSSVM model was shown to be capable of capturing the actual fluctuations of the OFS over time.In general,the overall modeling error(indicated by the RMSE index) of the LSSVM model was accepted and smaller than that of GA-Mamdani model.Conclusion The data-driven LSSVM modeling approach is effective for OFS estimation thanks to its superior generalization performance.