A Similarity Measure And Application Research For RSS Time Series

At present,how to effectively tap the potential information and knowledge of RSS data has attracted the attention of a large number of scientific researchers.However,the current RSS time series similarity measure analysis method is generally based on Euclidean distance,while the European distance can not capture the sequence on the time axis offset.Therefore,according to the shortcomings of RSS time series similarity measure,this thesis carries out relevant research work.Firstly,the limitations of the traditional similarity measure are analyzed,and the dynamic time regularization algorithm is used to solve the problem of RSS time series similarity of asynchronous time axis.In order to reduce the computational complexity,an improved two-step algorithm is proposed on the basis of sequence feature extraction and spectral clustering algorithm.Firstly,the sequence feature is extracted and the RSS time of similar motion trajectory is clustered by spectral clustering algorithm Sequence,and then in the cluster after each cluster with dynamic time regularization algorithm fine-grained detection of RSS time series similarity.In addition,the optimal clustering problem is solved according to the clustering validity index.Experiments show that this method can effectively solve the RSS time series similarity measure.Secondly,combined with the space-time invariance of convolution neural network,and can learn the advantages of low-dimensional spatial data mapping,the traditional similarity measure in the feature extraction and distance measurement learning the two steps in a framework,A method of RSS time series similarity measurement based on convolution neural network is proposed.Based on the K-NN classification algorithm,the feasibility of the proposed algorithm is verified by exposing the data set.Finally,the RSS time series similarity measure algorithm proposed in this thesis is applied to the uniqueness detection of mine entry personnel.The experimental results show that the proposed method can effectively measure the similarity between RSS time series and provide the theoretical basis for RSS data mining.At the same time,the similarity measurement algorithm based on RSS time series has a good practical application value,which has laid a good method and foundation for the subsequent engineering application.