PROACTIVE EVENT MONITORING BASED ON PREDICTIVE TIME SERIES ANALYSIS 

Kolesnikov Ilja Nikolaevich, junior java developer, LLC «KSK TECHNOLOGIES» (64 Suvorova street, Penza, Russia), E-mail: iljakolesnikoff@yandex.ru
Finogeev Alexey Germanovich, doctor of technical sciences, professor, sub-department of computer-aided design systems, Penza State University (40 Krasnaya street, Penza, Russia), E-mail:alexeyfinogeev@gmail.com 

338.26 

10.21685/2227-8486-2020-1-9 

Subject and goals. The article discusses methods of proactive monitoring of road transport infrastructure based on the collection and processing of big data about events on controlled sections of roads. In the process of monitoring, heterogeneous data is consolidated from many open sources and the characteristics of events are extracted in order to present their dynamics in the form of time series for analysis and prognostic modeling of the risks of emergencies and emergencies taking into account the influence of external factors.
Methods. To achieve the goal and objectives of the research, methods of collecting, consolidating and Big Data processing, identification, classification and clustering of events, a comparative analysis of the spectra of time series of their characteristics and influence factors are used. Big data on traffic accidents and incidents comes from distributed photo-radar photo and video recording complexes of offenses and vehicles, as well as from open sources on the Internet and from mobile means of communication of witnesses and participants in events.
Results and conclusions. Methods and tools were selected for collecting and analyzing big data about events in a proactive monitoring system, such as Hadoop, MapReduce and NoSQLе. The main methods for collecting and consolidating heterogeneous data for intelligent analysis and predictive modeling are considered. An algorithm for collecting and preparing textual data using the vector representation of words for machine learning a prediction system based on spectra of time series of events is presented. The results of proactive monitoring are necessary for a proactive response to possible negative events and incidents in the road environment to reduce emergency situations and provide emergency assistance. The proactive monitoring system under consideration is based on the methods of collecting and analyzing big data, and the intellectual analysis of data obtained from various information sources. The system proposes the use of prognostic modeling of risks of occurrence and development of events by constructing a spectrum of time series based on a vector representation of words. These methods will allow the proactive monitoring system to solve the problems of assessing and the risks of emergency and emergency events on the roads, taking into account the influence of external factors, to control road sections and vehicles, to track the locations of accident sites and other destructive events on the roads. 

big data, data mining, time series, predictive analytics, predictive modeling, Hadoop, MapReduce, NoSQL