基于机器学习的飞行训练数据集构建与应用  

Construction and application of flight training dataset based on machine learning

作　　者：路晶[1] 白一丹 钟一涛 陈燃 LU Jing;BAI Yidan;ZHONG Yitao;CHEN Ran(Civil Aviation Flight University of China,Guanghan,Sichuan 618307,China)

机构地区：[1]中国民用航空飞行学院,四川广汉618307

出　　处：《计算机应用文摘》2024年第16期76-79,共4页Chinese Journal of Computer Application

基　　金：民航飞行技术与飞行安全重点实验室飞行技术专题项目(FZ2022ZX52)。

摘　　要：高品质飞行训练数据集在推动科学发展中发挥着至关重要的作用。针对飞行训练领域中数据稀缺和缺乏高效质量增强方法的问题,围绕单发C172S飞机型号中的B-7943号飞机收集的数据,设计了一套基于机器学习的数据处理流程,旨在优化飞行训练数据的质量。该流程包括:利用滑动窗口修正数据中的极端异常值,应用先进的滤波技术平滑原始数据,以及使用随机森林算法填补缺失值。此外,通过灰度关联分析法评估了数据特征间的相关性。实验结果表明,该流程能够显著增强数据特征间的联系并有效提升数据质量。High quality flight training datasets play a crucial role in driving scientific development.Aiming at the problem of data scarcity and lack of efficient quality enhancement methods in the field of flight training,a machine learning based data processing flow was designed around the data collected from the B-7943 aircraft in the single engine C172S aircraft model,aiming to optimize the quality of flight training data.The process includes:using sliding windows to correct extreme outliers in the data,applying advanced filtering techniques to smooth the raw data,and using random forest algorithm to fill missing values.In addition,the correlation between data features is evaluated through grayscale correlation analysis.The experimental results show that this process can significantly enhance the connection between data features and effectively improve data quality.

关 键 词：机器学习 随机森林 滑动窗口 插补空值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]