【摘要】 民机驾驶舱显示界面可用性关系到飞行员满意度及飞行安全和效率，但目前该领域的可用性指标缺乏针对性。通过执行过程交互控制(executive process-interactive control, EPIC)模型分析获得民机驾驶舱显示界面可用性指标的维度：飞行任务、飞行员感官、飞行员认知和飞行交互，并在这4个维度的基础上进行提取和筛选。为解决小样本支持下的多维度指标数据集的筛选问题，使用基于层次聚类-粗糙集信息熵(HC-CEBARKNC)的指标聚类约简算法，并与K-means聚类-遗传算法进行了实例对比研究。运用支持向量机分类模型验证2种算法的性能和可靠性，试验结果表明：HC-CEBARKNC算法在指标聚类约简性能上表现更佳，能够有效简化民机驾驶舱显示界面的可用性指标体系、提炼可用性决策规则。更多还原

【Abstract】 The availability of civil aircraft cockpit display interface is related to pilot′s satisfaction as well as flight safety and efficiency, but current indicator system in this field lacks target-oriented focus. Firstly, in this paper, the flight mission, pilots′ senses, pilots′ cognition, and flight interaction dimensions are obtained by executive process interactive control model(EPIC), and extracted usability indicators from these dimensions. Secondly, to tackle the problem of filtering multidimensional indicator datasets supported by small samples, an indicator clustering reduction algorithm is proposed based on hierarchical clustering-rough set information entropy(HC-CEBARKNC) which compared to K-means clustering and genetic algorithm. Finally, the support vector machine(SVM) classification model was employed to verify performance and reliability of both algorithms. The experimental result shows the HC-CEBARKNC algorithm proposed has better evaluation accuracy that contribute to practical indicators reduction and decision rules screening.更多还原