摘要
为持续高效地学习不断产生的航班运行信息,提高航班延误预测模型学习新到达数据的效率,采用集成学习思想,提出了一种基于分类与回归树(classification and regression tree,CART)的增量学习算法.首先,将CART算法与Learn++算法结合实现了增量分类与回归树(incremental classification and regression tree,I-CART)算法;然后,进一步分析了基分类器间的区别和与精确度的关系,使用选择性集成算法来提高I-CART算法预测速率;最后,将该算法应用到航班延误预测中,增量地学习航班动态运行信息.实验结果表明,该算法有效地提高了模型预测效果.
To continuously and efficiently learn the constantly generated flight information and improve the efficiency of flight delay prediction model to learn new arrival data,an incremental learning algorithm based on classification and regression tree(CART)was proposed by using ensemble learning ideas.First,incremental classification and regression tree(I-CART)incremental learning algorithms were implemented by combining CART algorithm with Learn++algorithm.Then,based on the relationship between the difference and accuracy of basic classifiers,and the prediction rate of I-CART algorithm was improved by using the selective ensemble algorithm proposed in this paper.Finally,the incremental learning algorithm was applied to flight delay prediction.Results show that the incremental learning algorithm of flight dynamic information effectively improves the prediction performance of the model.
作者
王丹
王萌
王晓曦
杨萍
WANG Dan;WANG Meng;WANG Xiaoxi;YANG Ping(Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China;State Grid Management Institute,Beijing 102200,China)
出处
《北京工业大学学报》
CAS
CSCD
北大核心
2020年第11期1239-1245,共7页
Journal of Beijing University of Technology
基金
国家自然科学基金资助项目(61672505)。
关键词
航班延误
分类与回归树(CART)算法
增量学习
集成学习
选择性集成
机器学习
flight delay
classification and regression tree(CART)algorithm
incremental learning
ensemble learning
selective ensemble
machine learning
