guest ::
TY - THES AU - Chakrabarti, Arnab TI - Exploratory pipeline for analysis and visualization of large information spaces PB - RWTH Aachen University VL - Dissertation CY - Aachen M1 - RWTH-2023-05502 SP - 1 Online-Ressource : Illustrationen, Diagramme PY - 2023 N1 - Veröffentlicht auf dem Publikationsserver der RWTH Aachen University N1 - Dissertation, RWTH Aachen University, 2023 AB - The advent of the fourth industrial revolution (Industry 4.0) has fueled the capability of today’s scientific and commercial applications in generating, storing and processing massive amounts of data. Visual exploration of datasets having excessive data points and features leads to over-plotting and visual clutter, resulting in the loss of interpretability. This problem has been termed as “visual noise” in the field of high-dimensional data exploration. Our work in this thesis centers around the question of reducing visual noise and improving data interpretability in the quest of finding interesting patterns within high-dimensional datasets. This task is highly challenged by the so-called “curse of dimensionality”, which hinders visual exploration as a high number of data dimensions are required to be displayed on two-dimensional screen space. To mitigate such challenges, we present a comprehensive and flexible data exploration pipeline for handling high-dimensional data. We incorporated the proposed pipeline into the prototype “VizExplore”, which serves as an automated exploratory tool for monitoring production engineering processes. For the different stages of the pipeline, firstly, we propose dimensionality reduction methods for generating effective visualizations. With this, we address the open research problem of performing data reduction with the goal of minimizing information loss. Second, we propose a visualization ranking model. We develop novel methods of extracting visual characteristics from images and design a generic evaluation framework capable of comparing diverse visual structures in a uniform platform. Finally, we ingest the knowledge gathered from the first two layers and propose a visualization recommendation system using data-specific characteristics and user intended exploration tasks. We design recommendation engines using both rule-based and knowledge-based approaches and highlight the effectiveness of such recommenders in enhancing the functionalities of the exploratory process, especially in the field of high-dimensional data visualizations. We evaluated our exploratory tool in various settings, including data generated from engineering processes in the Internet of Production (IoP). We highlight the usability of our model in the IoP for handling high dimensional data and generating effective visualization recommendations, in turn, enhancing the decision-making capabilities for production systems. LB - PUB:(DE-HGF)11 DO - DOI:10.18154/RWTH-2023-05502 UR - https://publications.rwth-aachen.de/record/958959 ER -