Порівняння методів виявлення тренду на часових рядах малої довжини

Abstract

This work is devoted to the assessment of the criteria for determining the presence of a trend in short time series, developed on the basis of the k-means method and the method of determining the trend component using the discrete wavelet transformation, and comparing the performance of these methods. A software product was developed to test the hypothesis of no trend and compile static data on the performance of both methods. An analysis was performed to compare the performance of both methods to identify key differences between the methods.