Periodicity Recognition And Extraction Of Approximately Periodic Time Series

Study of periodic time series is a hotspot in related field nowadays. Currently in our daily life, there is a certain period features existing in time series of historical data in many events and phenomena. However the length of the period sometimes is unequal. Based on this consideration, the article is mainly about the study of not fixed cycle length of time series, i.e., the approximately periodic time series.First the paper introduces the concept of the approximately periodic time series and a key related method. Second, on this basis, we propose a new method to estimate the time transformation function-fitting estimation method. As extracting the data to fitting estimation method is very important, so this article compares two methods of estimating the data in the third section, and the result shows the second method is great. Last, we found that different sampling methods to identify the time series period is influential and for some sampling methods, time series period was unable to reflect its true cycle.By the way, to some extent, some methods in selecting data in estimation the thesis mentioned are influenced by the noise in time series. There is much space for improvement.