Visualization and Interaction for Knowledge Discovery in Simulation Data

Abstract

Discrete-event simulation is an established and popular technology for investigating the dynamic behavior of complex manufacturing and logistics systems. Besides traditional simulation studies that focus on single model aspects, data farming describes an approach for using the simulation model as a data generator for broad scale experimentation with a broader coverage of the system behavior. On top of that we developed a process called knowledge discovery in simulation data that enhances the data farming concept by using data mining methods for the data analysis. In order to uncover patterns and causal relationships in the model, a visually guided analysis then enables an exploratory data analysis. While our previous work mainly focused on the application of suitable data mining methods, we address suitable visualization and interaction methods in this paper. We present those in a conceptual framework followed by an exemplary demonstration in an academic case study.

A content analysis-based approach to explore simulation verification and identify its current challenges

“…Feldkamp, Bergmann [119] explore the role of knowledge discovery for online analysis of simulation data. They further explore the role that interactive visualization can play in supporting the knowledge discovery process by adding traceability to outcomes across experiment runs [120]. Time series plots are commonplace for displaying quantitative changes over time with a single parameter or feature represented by each line of the plot [121,122].…”

“…Radar charts convey time-oriented, multi-dimensional population data to convey trends and trajectories based on the input settings [29]. Parallel coordinate plots using color coded cluster representations have aided knowledge discovery by adding traceability to analyses involving multiple simulation runs' outcomes [120]. These techniques are valuable for revealing targeted insight into multivariate data, but they do not scale well as they result in large volumes of visual artifacts that require manual inspection.…”

“…Such occurrences may be the result of non-linear actions of subgroups or local networks [118,129], changing model structures over time [28], or the scale of networked or interconnected components [132]. Therefore, techniques need to differentiate between individual-level, subgroup-level, and population-level occurrences [60,120,133] to provide traceability between occurrences and model specifications. Network visualizations provide power for data integrations and transformations in revealing ways that can confirm hypotheses or identify new hypotheses by exploring the results of batches of experiments [125].…”

“…Numerous studies have utilized other sensory feedback to enhance the simulation experience [152][153][154], but they have not explored specific applications of use in supporting the verification process. Visual analytic approaches differentiate techniques' abilities to account for univariate, multivariate, special case, and temporal data types [120,124], but a deep dive on how and when specific techniques are applicable to existing verification challenge areas within simulations has not yet been conducted.…”

“…methods. Multivariate data can lead to challenges in conveying interpretable outcomes that clearly reflect temporal changes due to multiple dimensions, aggregations, or factor interactions [120,126]. Verification techniques need continued advancements for communicating complex analyses resulting from aggregations of simulation runs and interaction points hidden within layers of interacting components within the simulation.…”