ANOMALIES DETECTION BASED ON HETEROGENEOUS DATA IN DISTRIBUTED MONITORING SYSTEMS

UDC 004.056
DOI: 10.36871/2618-9976.2020.11.002

Authors

Bekeneva Yana A.
Candidate of Technical Sciences, Saint Petersburg Electrotechnical University «LETI», Department of Computer Science and Engineering, Saint Petersburg, Russia

Abstract

Various kinds of abnormal situations in the processes can be associated with both minor deviations and serious malfunctions or violations that can lead to irreparable consequences and financial losses. Timely identification of anomalies allows you to influence the course of the process and minimize the consequences of detected deviations from the normal course of the process. Anomaly detection is one of the tasks of data analysis. Modern monitoring systems contain a large number of different devices, the data from which can be used as initial data for intellectual analysis. Data preprocessing has a great influence on the quality of the analysis. The paper presents the options for using various methods of data analysis to solve the problems of detecting anomalies in the processes associated with the movements of moving objects. The stages of data preparation for different methods of analysis are described. Much attention is paid to the increasingly popular methods of intelligent analysis of processes. The features of the input data format for the methods of mining processes are considered, as well as the features of the assignment of attributes. The proposed data processing methods were tested on several datasets related to the movements of trucks in the distributed territory of the organization and the movements of employees in an office building

Keywords

Data mining
Data preparation
Distributed monitoring systems
Process analysis
Anomaly detection
Event logs