The Research Of Semi-Supervised Classification On Time Series

Time series data refers to a set of data arranged in order of time,which is widely used in various fields.Most existing time series classification methods require a large number of labeled samples,but such data is very difficult to obtain and expensive.Therefore,when the number of labeled samples is small,how to use a large number of unlabeled samples to improve the classification performance has become a problem of great concern.Establishing an accurate classifier requires a large amount of data with category labels.However,in real-world applications,there are a large number of unlabeled data,and labeled data is often difficult to obtain.Time series semi-supervised classification uses a small amount of labeled data and a large amount of unlabeled data to construct a classifier with higher accuracy.It reduces the inconvenience caused by factors such as manual marking and is difficult to obtain marking data.This paper makes innovations on semi-supervised classification methods for univariate and multivariate time series.The main research work is as follows:(1)For the semi-supervised classification of univariate time series,a semi-supervised classification algorithm based on Locality Preserving Projections and a semi-supervised classification algorithm based on Piecewise Aggregate Approximation are proposed.Most of the existing researches directly conduct semi-supervised classification of the original time series.In general,the dimension(length)of the time series is relatively high,and it is very important to select the appropriate dimensionality reduction technique in the semi-supervised classification method.In this paper,a time-semi-supervised classification method LPP_SSCTS based on Locality Preserving Projections is proposed.LPP_SSCTS can solve the curse of dimensionality and eliminate noise while selecting the appropriate parameters,and also enable the reduced-dimensional data to clearly maintain the local neighborhood information of the original data.The experimental results on 15 time series datasets show that the proposed algorithm is significantly better than the existing ones.In the subsequent research,this paper proposes a time series semi-supervised classification algorithm PAA_SSCTS based on Piecewise Aggregate Approximation.The method first uses the Piecewise Aggregate Approximation to reduce the dimension of the time series to the time series samples,and then semi-supervised the data after the dimension reduction.Experimental results in different time series data sets show that the classification performance of this method is significantly better than the existing methods.(2)For the semi-supervised classification of multivariate time series,this paper proposes a semi-supervised MTS classification method based on Two-Dimensional Singular value Decomposition,due to the complex relationship between multivariate time series(MTS)variables,there is less research on semi-supervised classification of MTS.This method first computes the eigenvectors of row-row and column-column covariance matrices,and then extracts feature matrices from MTS samples.The number of rows and columns of the feature matrix is not only lower than the original MTS sample,but also clearly considers the two-dimensional nature of the MTS sample.The experimental results on 10 MTS datasets show that the semi-supervised classification performance of this method is significantly better than the method using Extended Frobenius norm,Center Sequence and based on One Dimensional Singular Value Decomposition.