Towards an Ontology for Representing Time Series Knowledge: Motivation, Requirements and Concept

Time series analysis is an essential task in a variety of domains, where specialized ontologies can be leveraged to define data- and analysis-related specificities. However, the data management landscape lacks a comprehensive semantic data model for representing time series insights produced by data analysis operations. As these insights differ in format and interpretation, data reusability as well as the potential for synergy effects across various analysis methods are limited by the absence of such standardized data models. In this paper, we introduce an ontology designed to systematically categorize insights inferred from time series analytics. By representing time series characteristics, such as anomalies, trends, or motifs as concrete knowledge entities, our approach facilitates the reuse and exploitation of knowledge across different levels of abstraction. To further enrich and combine this knowledge with prior information, it enables the integration and association of domain-specific facts provided by additional resources including human experts. In addition to introducing the ontology concept including requirements and limitations, we outline the applicability of our proposal with the help of an example.