Event Series Generation and Analysis Based on Multifaceted Ontology

Abstract

The article presents an approach to the analyzing processes in different domains using data from various Internet sources (open databases, news feeds, social networks, etc.). This one is suitable to carry out cross-disciplinary research encompassing processes in various fields (for example, economics, medicine, politics, ecology, etc.) in which events can have mutual affects. The concept of event series is given as the main one in this study. Event series are defined via analogy with time series as a collection of values of some parameters of the investigated processes, where the type of event is indicated instead of the measurement time. The event series analysis should take into account not only the relationship of measurements with time (or the events chronology), but also the causal relationships that can be identified at the study of processes. The event series formation is based on using multifaceted ontology describing different aspects of research such as data sources and structure of information extracted from them to solve user’s tasks, as well as domains and events, rules that characterize events, and the methods to solve tasks. The ontology is used when working with unstructured data to build an event log, which is formed in the first step when constructing event series. Next, the ontology is used to pre-process data before performing process mining tools applied to creating and analyzing models of processes. Using multifaceted ontology experts can define new rules for pre-processing data and generating event logs based on the concept of event-time series. These tools allow to generate more informative models.