An Improved Wavelet-Time Series Method And Its Application

The wavelet analysis method is one of time-frequency processing methods with superior performance,which can retain the time domain information of the signal based on frequency domain analysis,so it has attracted widespread attention.However,there is a kind of data in nature that varies significantly with time and seasonal differences,such as ozone data with highly non-stationary.The wavelet analysis method with different functions and parameters will greatly affect the stability of the algorithm,so using the same wavelet function for all data cannot achieve the maximum reconstruction.In view of this defect of wavelet analysis method,this paper designs an improved wavelet analysis method based on the particularity of ozone concentration data.This method is used to decompose and reconstruct the data,and the atmospheric ozone concentration is predicted combined with the time series prediction model.The new method groups the data according to the characteristics,which is similar to the idea of adaptive function.Then different wavelet functions and decomposition levels are applied to each set of data before the results are integrated.The results show that compared with the wavelet decomposition method of a single function,the multiple adaptive wavelet analysis method has higher goodness-of-fit and smaller error.Then,combined with the time series method,the ozone concentration was predicted,and an improved wavelet-time series prediction model was constructed.Through comparison and analysis,the improved model proposed in this paper can handle non-stationary and highly variable time series better,make the model more reliable,and obtain more ideal prediction results.