基于AdaBoostSVM算法的时间序列分类方法研究

Research on A Time Series Classification Method Based on AdaBoostSVM Algorithm

时间序列数据广泛应用于各大领域,传统的时间序列数据分类方法存在精准度低、错误分类等问题。为了提升时间序列数据分类的精准性及稳定性,提出了基于AdaBoost和SVM级联算法的时间序列数据分类方法,并针对16类UCR时间序列数据进行实验分析。实验结果表明,AdaBoostSVM算法模型平均分类精准性达96.35%,较传统的1-NN等分类方法高5%,较LSTM深度学习算法分类精准度高21%,精准性更高,稳定性更优。

Time series data is widely used in various fields.Traditional time series data classification methods have problems such as low accuracy and misclassification.In order to improve the accuracy and stability of time series data classification,a time series data classification method based on AdaBoost and SVM cascade algorithm is proposed.Experiments are conducted on 16 types of UCR time series data.The experimental results show that the AdaBoostSVM algorithm model achieves an average classification accuracy of 96.3%,which is 5%higher than traditional classification methods such as 1-NN,and 21%higher than the LSTM deep learning algorithm.The accuracy is higher and the stability is better.