Finding Anomalies in the Time Series Data by Using the Wave Equivalence Method

This paper is dedicated to the problem of detecting anomalies in the particular time series data, which consists of wave-like data. Anomaly detection is implemented in various fields and can be used in many automated control systems. Wave-like data is characterized by such attribute as seasonality, and every data season is a wave that has a similar rise, top limit, and fall of data. In this case, detecting when the data have waves with different height and width is necessary. Setting only the usual limits by setting the minimum and maximum value on one axis and checking whether the data goes beyond the height limits is ineffective because, in this case, it does not take into account the data on the wavelength. Also, it is necessary to set different limits every time for different time series. Thus, it is necessary to develop an algorithm that can automatically separate waves, calculate the base characteristics of these waves, and, based on the conducted calculations, detect waves that do not correspond with the regular wave parameters. The solution to identifying anomalies in the particular type of time series data is a new method based on derivatives that can detect a rise, top limit, and fall of waves, split data by waves, and check the wave equivalence.