Minimum error estimation using wavelet for time series similarity search
对时间序列查询的最优小波误差估计

Wavelet transform is used as a dimensionality reduction method to permit efficient similarity search over high-dimensional time series data. The traditional algorithms use the first k wavelet coefficients as an approximation of the original time series. But sometimes choosing the first k coefficients is not the best when approximating the original time series. Perhaps choosing other k wavelet coefficients is better than choosing the first k. The theorem was given to better approximate the original time series set. The main idea of the theorem was to choose the k columns of the wavelet sequences set which had the maximum square sum. The experiments show that it can better reduce the relative error compared with original algorithms.