Research On Multivariate Time Series Of Cultural Relics Based On Wavelet Theory

Immovable cultural relics have high historical value,cultural value and scientific value,as an important carrier of ancient Chinese civilization.Due to changes in the natural environment and other reasons,immovable cultural relics have different disease,and its state of preservation is greatly threatened.With the application of Internet of Things in the protection of immovable cultural relics,although there are many problems are avoided in the means of traditional ways,such as the not timely response to abnormal changes,the incomplete obtaining information and low efficiency,it is very necessary to use the data of immovable cultural relics to analyze and forecast cultural relics itself and the risk factors,which becomes the key to the next step in the information protection.Based on the analysis of the characteristics of data from Internet of Thing about immovable cultural relics,this thesis studies the correlation analysis method and forecast modeling method based on wavelet theory,and designs the disease forecast and analysis system of cultural relics based on the above method.The main work is as follows:(1)Multivariate time series correlation analysis of cultural relics.For the limitations of wavelet correlation analysis theory about only for the unit time series,and the importance of quantitative delay-correlation analysis of cultural relics multivariate time series,it is proposed a correlation analysis method of multivariate time series based on wavelet theory.Firstly,it combines the defined wavelet correlation coefficient with the stepwise regression to realize the reduction of the variables.Secondly,the wavelet correlation coefficient is obtained by constructing the wavelet delay-correlation coefficient formula.And the quantitative result is described.It is shown that this method is able to enhance the applicability of wavelet correlation analysis of multivariate time series,and obtain the quantitative correlation and delay-correlation of multivariate time series.(2)Multivariate time series forecast modeling of cultural relics.In view of the complex correlation of multivariate and the lacked factor in the forecast of multivariate time series of cultural relics,it is proposed an improved forecast modeling method,including the variable selection method based on wavelet correlation theory and delay-correlation,as the reference factor of forecast modeling.Firstly,the variable selection method is designed according to the wavelet correlation analysis method and the minimum correlation criterion.The characteristic variables are selected effectively.And the forecast model based on BP neural network is established,with the quantitative delay-correlation and the selected characteristic variables.It is shown that this method could further improve the accuracy of the forecast model of multivariate time series.(3)Disease forecast and analysis system of cultural relics.For the missing feature of data processing in the existing Internet of Things of immovable cultural relics,it is designed a disease forecast and analysis system of cultural relics in which integrate above analysis algorithms.The system not only provides a scientific and reliable evidence,but improves the application system of immovable cultural relics,which plays an important role in the preventive protection of cultural relics.