时序数据挖掘的预处理研究综述  

Summary of research on preprocessing of time series data mining

作　　者：张友浩 赵鸣[1] 徐梦瑶 赵光晶 ZHANG Youhao;ZHAO Ming;XU Mengyao;ZHAO Guangjing(School of Air Transport,Shanghai University of Engineering Science,Shanghai 201620,China)

机构地区：[1]上海工程技术大学航空运输学院,上海201620

出　　处：《智能计算机与应用》2021年第1期74-78,共5页Intelligent Computer and Applications

摘　　要：由于时序数据的海量、高维等特点,对原始数据进行预处理是实现时序数据挖掘的重要步骤。有效的预处理不仅能去除噪声、冗余,减低数据规模、提升计算效率,还能通过提取数据特征,揭示数据关系,有助于实施时序数据挖掘。本文从数据缺失、数据噪声、离群点/异常值、特征提取和模式表示几个方面,对已有的原理和方法进行分析归纳,为后续时序数据挖掘研究提供参考。Due to the massive and high-dimensional characteristics of time series data,preprocessing the original data is an important step for time series data mining.Effective preprocessing can not only remove noise and redundancy,reduce the scale of data,and improve computing efficiency,but it can also reveal data relationships by extracting data features,which helps to implement time series data mining.This paper analyzes and summarizes the existing principles and methods from the aspects of data missing,data noise,outliers/anomalous value,feature extraction and pattern representation,and provides references for subsequent time series data mining research.

关 键 词：时间序列 预处理 数据清洗 特征提取 模式表示 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]