RESEARCH AND DEVELOPMENT OF A METHOD TO SELECT THE SEGMENTATION ALGORITHM FOR THE PROBLEM OF DETECTING IRREGULARITIES IN TIME SERIES DATA

Research and improvement of the application segmentation method for the problem of detecting anomalies in time series data aims to optimize the process of identifying abnormal events. This method mainly focuses on using segmentation techniques to divide time series data into sub-segments, combined with machine learning models, especially deep learning models, to automatically extract features and identify abnormal manifestations. Thereby, the model is optimized for real-time application and has the ability to flexibly adapt to fluctuations in time series data. At the same time, choosing the appropriate error to segment the time series to minimize prediction errors and increase reliability. The results of the method are evaluated and compared on test data sets, and its practical application is verified in system monitoring, fault prediction, and resource management. Propose improvements to enhance the performance and practical application of the method, aiming to provide a reliable solution to the problem of anomaly detection in time series data.