基于改进KNN算法的飞行员工作负荷评估

Pilot Workload Assessment Based on Improved KNN Algorithms

针对使用主观量表评估飞行员工作负荷易受主观因素干扰的问题,将飞行员工作负荷评估试验划分成连续的15 s时间窗口,基于时间窗口内的客观绩效和生理数据,建立飞行员工作负荷评估模型。使用插值、去均值、归一化等方法预处理数据后,再将生理数据变化量作为工作负荷的特征维度,并基于生理数据变化量改进KNN算法,对工作负荷进行分类。通过引入生理数据变化量作为工作负荷的特征维度,优化分类模型的数据结构后,各传统分类算法的测试集F1分数均得到提高;使用生理数据变化量改进KNN算法后,高负荷数据分类准确率达到71%,总体准确率能达到88.5%;相比于传统KNN算法,高负荷数据分类准确率提升36.5%,总体准确率提升6.3%。

To address the problem that subjective scale assessment of pilot workload is susceptible to subjective factors,this paper divided pilot workload evaluation test into continuous 15 seconds time window,and established pilot workload evaluation model based on objective performance and physiological data in the time window.After preprocessing the data by interpolation,de mean and normalization,the change of physiological data is used as the feature dimension of workload,and KNN algorithm is improved based on the change of physiological data to classify workload.After the introduction of physiological data variation as the feature dimension of workload to optimize the data structure of classification model,the F1 scores of test sets of traditional classification algorithms are improved;after using physiological data changes to improve KNN algorithm,the classification accuracy of high load data is 71%,and the overall accuracy is 88.5%.Compared with the traditional KNN algorithm,the classification accuracy of high load data is increased by 36.5%,and the overall accuracy is increased by 6.3%.