TU Darmstadt / ULB / TUbiblio

Implementing Clustering and Classification Approaches for Big Data with MATLAB

Pitz, Katrin and Anderl, Reiner (2019):
Implementing Clustering and Classification Approaches for Big Data with MATLAB.
In: Proceedings of the Future Technologies Conference (FTC) 2018, Cham, Springer Nature Switzerland AG, pp. 458-480, [Book Section]

Abstract

Data sets grow rapidly, driven by increasing storage capacities as well as by the wish to equip more and more devices with sensors and connectivity. In mechanical engineering Big Data offers new possibilities to gain knowledge from existing data for product design, manufacturing, maintenance and failure prevention. Typical interests when analyzing Big Data are the identification of clusters, the assignment to classes or the development of regression models for prediction. This paper assesses various Big Data approaches and chooses adequate clustering and classification solutions for a data set of simulated jet engine signals and life spans. These solutions include k-means clustering, linear discriminant analysis and neural networks. MATLAB is chosen as the programming environment for implementation because of its dissemination in engineering disciplines. The suitability of MATLAB as a tool for Big Data analysis is to be evaluated. The results of all applied clustering and classification approaches are discussed and prospects for further adaption and transferability to other scenarios are pointed out.

Item Type: Book Section
Erschienen: 2019
Creators: Pitz, Katrin and Anderl, Reiner
Title: Implementing Clustering and Classification Approaches for Big Data with MATLAB
Language: English
Abstract:

Data sets grow rapidly, driven by increasing storage capacities as well as by the wish to equip more and more devices with sensors and connectivity. In mechanical engineering Big Data offers new possibilities to gain knowledge from existing data for product design, manufacturing, maintenance and failure prevention. Typical interests when analyzing Big Data are the identification of clusters, the assignment to classes or the development of regression models for prediction. This paper assesses various Big Data approaches and chooses adequate clustering and classification solutions for a data set of simulated jet engine signals and life spans. These solutions include k-means clustering, linear discriminant analysis and neural networks. MATLAB is chosen as the programming environment for implementation because of its dissemination in engineering disciplines. The suitability of MATLAB as a tool for Big Data analysis is to be evaluated. The results of all applied clustering and classification approaches are discussed and prospects for further adaption and transferability to other scenarios are pointed out.

Title of Book: Proceedings of the Future Technologies Conference (FTC) 2018
Series Name: Advances in Intelligent Systems and Computing
Volume: Volume 1
Place of Publication: Cham
Publisher: Springer Nature Switzerland AG
Uncontrolled Keywords: Big Data, clustering, classification, k-means, discriminant analysis, neural networks, MATLAB
Divisions: 16 Department of Mechanical Engineering
16 Department of Mechanical Engineering > Department of Computer Integrated Design (DiK)
Event Title: Future Technologies Conference (FTC) 2018
Event Location: Vancouver, BC
Event Dates: 13.-14.11.2018
Date Deposited: 04 Jan 2019 08:21
Alternative keywords:
Alternative keywordsLanguage
Big Data, Cluster, Klassen, k-Means, Diskriminanzanalyse, neuronale Netze, MATLABGerman
Alternative Abstract:
Alternative abstract Language
Datenmengen nehmen beständig zu, getrieben durch wachsende Speicherkapazitäten sowie durch den Wunsch, immer mehr Geräte mit Sensoren und Konnektivität auszustatten. Im Maschinenbau bietet Big Data neue Möglichkeiten, um aus vorhandenen Daten Wissen für Produktdesign, Fertigung, Instandhaltung und Versagensvermeidung zu gewinnen. Typische Interessen bei der Analyse von Big Data sind das Identifizieren von Clustern, das Bestimmen von Klassenzugehörigkeiten oder die Entwicklung von Regressionsmodellen zu Vorhersagezwecken. In diesem Paper werden verschiedene Big-Data-Ansätze bewertet und geeignete Cluster- und Klassifizierungslösungen für einen Datensatz simulierter Triebwerkssignale und -lebensdauerinformationen gewählt. Diese Lösungen umfassen k-Means-Clustering, lineare Diskriminanzanalyse und neuronale Netze. MATLAB wird aufgrund seiner Verbreitung in den Ingenieurdisziplinen als Programmierumgebung für die Implementierung gewählt. Die Eignung von MATLAB als Werkzeug für die Big Data-Analyse wird beleuchtet. Die Ergebnisse aller angewandten Clustering- und Klassifizierungsansätze werden diskutiert und Perspektiven für weitere Anpassungs- und Übertragungsmöglichkeiten auf andere Szenarien aufgezeigt.German
Export:
Suche nach Titel in: TUfind oder in Google

Optionen (nur für Redakteure)

View Item View Item