Digital Library of Gesellschaft für Informatik e.V.
Not a member yet
40758 research outputs found
Sort by
Relationship Discovery for Heterogeneous Time Series Integration: A Comparative Analysis for Industrial and Building Data
No full textCyber-physical systems like buildings and power plants are monitored with ever-increasing numbers of sensors, gathering massive and heterogeneous time-series datasets collected in data lakes. Appropriate meta-data, describing both the function and location of each sensor, is essential for any profitable use of the data but is often not available or incomplete. Particularly, information about related sensors, meaning sensors belonging to the same functional subsystem, might be hard to derive if appropriate meta-data is unavailable. While various approaches exist for automatic meta-data extraction from relational databases, the unique characteristics of heterogeneous time-series data necessitate specialized algorithms. Among the general algorithms developed for time-series meta-data inference, only a few are concerned with relationship discovery despite the critical importance of this information in many meta-data formats. Nevertheless, other domains offer a variety of measures for pairwise relationship discovery in homogeneous time-series collections. This paper consolidates these measures and evaluates their performance for identifying related but heterogeneous time series from the same functional subsystem within industrial facilities. We evaluate the methods on a collection of different datasets to extract promising relationship measures from the literature and show that there are other better-performing candidates than the common Pearson Correlation Coefficient
Plaquette: Visualizing Redundancies in Relational Data
No full textFunctional dependencies are a fundamental topic in database education, yet students often find the concept abstract. To foster understanding of functional dependencies that capture redundancies in relational data, we propose an educational software tool called Plaquette. This tool offers a novel visualization of redundancies by coloring cells that contain redundant data, where deeper hues indicate stronger redundancies. In analogy to plaque tests at the dentist's office, we refer to these redundancies as "plaque". In our demo, we begin by recapping the definition of functional dependencies, followed by introducing our concept of plaque. Demo participants are then invited to interact with Plaquette, and assess sample scenarios that showcase small examples of relations and functional dependencies. By swiping through scenarios with correct but also fake plaque on a tablet in Tinder-style interaction, participants can playfully improve their intuition for the concept of functional dependencies. Beyond contributing a new teaching tool, our ultimate goal is to assist data analysts explore redundancies in real-world data, using tools like Plaquette
Operational design domains and dynamic driving tasks for both off- and on-road agricultural use cases
No full textThe integration of automation in agricultural machinery and robots promises to increase efficiency, productivity, and safety in agriculture, mainly as machines navigate between diverse environments such as fields and public roads. This paper examines how to effectively combine ISO 18497 and SAE J3016 standards to achieve a development framework for autonomous agricultural machines. ISO 18497 provides comprehensive guidelines on off-road applications, ensuring safe and precise operations in dynamic field environments. In contrast, SAE J3016 offers a well-structured framework of operational design domain (ODD) and dynamic driving tasks (DDT) for on-road vehicular automation. By analyzing these standards, the research aims to establish a consistent point of view on safety and performance that simplifies development and enhances functionality across various operational scenarios
Identifizierung von Finanzierungs- und Trägerschaftsmodellen für das LeguDash-Dashboard durch eine LLM-gestützte Analyse qualitativer Experteninterviews
No full textIn Nischenmärkten wie dem europäischen Leguminosenmarkt ist Markttransparenz entscheidend, um wettbewerbsfähige Bedingungen zu fördern und fundierte Vermarktungsentscheidungen zu ermöglichen. Das LeguDash-Dashboard soll diese Transparenz durch eine öffentlich zugängliche und interaktive Informationsplattform erhöhen. Um mögliche Trägerschaftsmodelle zu diskutieren, wurden Interviews mit Expert:innen durchgeführt. Die Interviews wurden durch qualitative Analysen sowohl manuell als auch mithilfe einer Large Language Model (LLM)-gestützten inhaltsanalytischen Methodik ausgewertet. Für die LLM-gestützte Analyse wurden ChatGPT und der AI Assist von MAXQDA verwendet. Eine weitere Auswertung mit dem Open Source Sprachmodell LlaMA-13b ist noch in Arbeit. Im Ergebnis der Interviews wurden verschiedene Finanzierungsmodelle vorgeschlagen, darunter öffentliche Finanzierung, Abonnement-basierte Bezahlinhalte mit einer „Paywall“ und ein Modell, bei dem Unternehmen Daten im Austausch für kostenlosen Zugang bereitstellen. Flexible, anpassbare und in bestehende Systeme integrierbare Lösungen werden gefordert. Methodisch zeigte die LLM-gestützte inhaltsanalytische Auswertung des qualitativen Datenmaterials erste nützliche Ansätze, jedoch gab es einige Unterschiede zur manuellen Analyse. Unter anderem wurden wichtige Aspekte von LLMs teilweise ausgelassen oder verkürzt, was zur Folge hatte, dass Lücken in der Auswertung auftauchten. Des Weiteren wurden einige Sätze kontextual fehlinterpretiert. Hier zeigte sich, dass die durch Fachwissen gestützte manuelle Auswertung wesentlich präziser und aussagekräftiger war, da sie eine tiefere und kontextgerechtere Interpretation der Daten ermöglichte
CheDDaR: Checking Data — Data Quality Report
No full textIn the data-driven world, high data quality is critical for effective decision making and advanced applications such as machine learning. Data often suffer from low quality, necessitating thorough cleaning. In this work, we introduce CheDDaR, a framework for detecting data quality issues early with minimal effort using various data quality metrics and multiple verification methods. The goal of CheDDaR is to keep manual effort low while maximizing the scope of the test. While CheDDaR is extensible, it already comes with ten metrics and five verification methods, which serve to generate precise reports highlighting data quality problems. Its effectiveness is demonstrated through a test dataset, showing the framework’s capability to identify and assess data quality issues effectively when combining multiple verification methods
Variability Modeling of Products, Processes, and Resources in Cyber-Physical Production Systems Engineering
No full textThis summary of a paper (with the same title) initially published in the Journal on Systems and Software (JSS) discusses efficient variability exploration of products, production processes, and production resources in engineering Cyber-Physical Production Systems
Formal Synthesis of Uncertainty Reduction Controllers
No full textIn this extended abstract, we summarise our work on the Parley approach and uncertainty reduction controllers. This work has been published under the title “Formal Synthesis of Uncertainty Reduction Controllers” at the Conference on Software Engineering for Adaptive and Self-Managing Systems (SEAMS’24)
Deep-learning-based quantification of sow activity during farrowing
No full textThe activity of sows after farrowing significantly influences the occurrence of crushing losses. For this reason, sow activity is often quantified in scientific research through video analysis. To improve practicality, this study developed an automated approach based on the YOLOv10 object detection framework, addressing the big data challenge. The core idea of this approach was to calculate an activity score based on the number of postural changes. Therefore, the model detects the sow’s posture for each frame in the videos. The raw data from the model is then filtered, and the number of postural changes is calculated, from which an activity score is derived for each video. For comparison with a human-generated activity assessment, the videos were finally categorized into three classes based on the activity score. The final performance evaluation across activity classes shows an accuracy of 0.945. Based on these results, it can be concluded that this approach provides an automated alternative to manual video analysis for determining sow activity
Robust Plan Selection using Cardinality Distribution Models
No full textEfficient query optimization requires an accurate estimation of cardinalities. Various techniques to improve cardinality estimation all share a common limitation: by classic definition, cardinality estimates are single-point values that drive the generation and selection of candidate execution plans. Relying on a wrong estimate can significantly impact the efficiency of query execution and therefore overall database performance. Rarely has the uncertainty of estimators been assessed or incorporated into the plan selection. Rather than solely relying on a single value estimate, we seek to quantify the risk associated with using each estimate, aiming to avoid those suspected to be wrong, unreliable or generally not trustworthy. To this end, we reformulate the cardinality estimation problem, introduce a novel class of learned cardinality estimators fit for this refined formulation, and explore their application in a robust plan selection strategy by implementing a simple query optimizer