An Ontological Approach to Modelling and Analysing System Performance

点击翻译按钮获取中文摘要

This research focuses on one aspect of the design and analysis of integrated systems, namely the formalization, standardization and automated analysis of system performance measurement (referred to as Metrics from here on in). One class of systems we wish to study is cities and the performance measures we wish to formalize and analyze are "city indicators." City indicators measure city services and functions, including Education, Health, Safety, Transportation, and Water. "There are thousands of different sets of city (or urban) indicators and hundreds of agencies compiling and reviewing them. . However, these indicators are usually not standardized, consistent or comparable (over time or across cities) ..." (Hornweeg et al., 2007). This problem has been exacerbated by the "Open Data" movement where cities are publishing vast amounts of data, but no two cities nor departments within a city use the same data models (Fox, 2013a). A consequence of this "babel" of open data is that it is difficult to combine, analyze and compare city data. The problem does not end with an agreement on the surface definition of an indicator, as each term within the definition has to be further defined (i.e., grounded) and supporting data provided.**We believe it is possible to identify and create a generic, reusable ontology for Metrics. We believe an ontology can be created by combining existing foundation, heterogeneous ontologies, and extending them with trans-foundation axioms. By heterogeneous, we mean ontologies that span analytical models (e.g., ratios), statistical models (e.g., population measurements), geo-spatial models (e.g., city boundaries), temporal models (e.g., time periods) and description logic models (e.g., definitions of students, teachers, etc.). But also incorporate meta-knowledge such as provenance, validity and trust. We believe that the ontology will make it possible to automate the drilling down and detection of inconsistencies in the supporting, heterogeneous data. Hence, our first objective is to forma