Article

Article title THE METHOD OF MULTISTEP PREDICTION OF ANOMALIES IN TEMPORAL DATA
Authors S.M. Kovalev
Section SECTION III. ARTIFICIAL INTELLECT AND INDISTINCT SYSTEMS
Month, Year 07, 2013 @en
Index UDC 519.816
DOI
Abstract In this paper, we develop a new method for the detection of anomalies in the temporal data on the basis of a multi-step methods for prediction. The proposed method is based on the analysis of the dynamics of the probability values of the anomaly on the input model.  As a means of calculating the probability values of the anomaly, we suggest using truth functions of stochastic Markov models with income. We prove a theorem which is the rationale for the use of the method. To convert to a real process of Markov, we propose to use an adaptive fuzzy system that can show a set of point states of the initial process on a granular scale fuzzy integral feature. To convert to a real process of Markov, we propose to use an adaptive fuzzy system that can show a set of point states of the initial process on a granular scale fuzzy integral feature.

Download PDF

Keywords Temporal data; anomaly detection; Markov model; temporal-difference learning; adaptive fuzzy system.
References Yeung D.Y., Ding Y.X. Host-based intrusion detection using dynamic and static behavioral models, Pattern Recognition. – 2003. – № 36. – С. 229-243.
2 Ковалев С.М. Упреждающее распознавание нечетких темпоральных паттернов в потоковых данных // Тринадцатая национальная конференция по искусственному интеллекту с международным участием КИИ-2012: Тр. конференции. Т. 2. – М.: Физматлит, 2012. – С. 313-322.
3 Курейчик В.В., Курейчик В.М., Сороколетов П.В. Анализ и обзор моделей эволюции. Известия РАН // Теория и системы управления. – 2007. – № 5.
4 Sutton R. Learning to predict by the method of temporal differences // Machine Learning. - 1988. – № 3 (1). – С. 9-44.
5 Малинейкий Г.Г., Потапов А.Б. Русла и джокеры: о новых методах прогноза поведения сложных систем // Препринт ИМП ИИ. им. М.В. Келдыша РАН. 2001.

Comments are closed.