摘要
采用当前算法挖掘时间序列数据时,不能有效的降低噪声对时间序列数据挖掘过程造成的影响,不能通过提高集群规模缩短数据处理时间,存在加速比低和可扩展性差的问题。将多目标决策理论应用到时间序列数据挖掘中,提出基于多目标决策的时间序列数据挖掘算法,挖掘时间序列之间对时间序列进行预处理,消除时间序列中存在的噪声,提取时间序列中存在区域极值点,对提取得到的区域极值点做等长处理,获得由极值点构成的序列。通过多目标决策方法利用获取的极值点构建决策矩阵,对比决策对象之间存在的差值,采用偏好函数将差值转变为对应的偏好度,对偏好度进行排序,根据排序结果对时间序列数据做聚类处理,实现时间序列的数据挖掘。仿真结果表明,所提算法的加速比高、可扩展性好。
In current algorithms, the influence of noise on the time series data mining process cannot be effectively reduced. Therefore, a time series data mining algorithm based on multi-objective decision was proposed. Firstly, the time series were preprocessed before mining time series, so as to eliminate the noise existing in time series and extract the regional extreme points from the time series. Secondly, the extracted regional extreme points should be of equal length, so that the sequence consisting of extreme points was obtained. Then, the multi-objective decision-making method was used to construct the decision matrix by those extreme points. In addition, the difference values between the decision objects were compared. The preference function was used to transform the difference value into the corresponding preference degree. Meanwhile, the preference degrees were sorted. According to the sorting result, the data of time series were clustered and thus to realize the data mining of time series. Simulation results show that the proposed algorithm has higher acceleration ratio and better scalability.
作者
何保荣
HE Bao-rong(Software Institute,Henan University of Animal Husbandry and Economy,Zhengzhou Henan 450046,China)
出处
《计算机仿真》
北大核心
2019年第11期243-246,共4页
Computer Simulation
关键词
多目标决策
时间序列
数据挖掘
Multi-objective decision making
Time series
Data mining