文件名称：A New Approach for Time Series Decomposition and Prediction (ITISE 2017)
文件大小：70KB
文件格式：DOCX
更新时间：2021-01-12 04:27:32
time series, decompose, predict
Abstract: We proposed an approach to decompose a univariate time series into a set of sine functions whose frequencies are not necessarily integers as the dis-crete Fourier transform, and designed an efficient algorithm of linear time com-plexity to calculate the parameters of the sine functions. We also used such a decomposition to make predictions into the future, and showed the prediction accuracy is comparable with other approaches especially in cases of quasi-periodic time series.