AAL Data Cluster Analysis. Theory and Implementation

Inhaltsverzeichnis (Table of Contents)

• Introduction
  • Motivation
  • Outline
  • The Goal
  • Methodology
• Theory Section
  • Gaussian Mixture Model
  • Dirichlet Process GMM
  • Hierarchical Clustering
• Implementation
  • Software Specification
  • Data and Sensors
  • Data Visualisation and Inspection
  • Sensor values Discretisation and Extraction
  • Unsupervised Event Extraction
  • Data Structure for Event Analysis
  • Data Quality and Quantity
  • Predictive Analysis
    • Hourly Binning Analysis
    • Clustering Analysis
      • Hierarchical Clustering Analysis
      • Dirichlet Process Gaussian Mixture Model Clustering Analysis
• Summary
  • Results
  • Discussion

Zielsetzung und Themenschwerpunkte (Objectives and Key Themes)

This thesis explores the potential of unsupervised prediction and clustering techniques for analyzing user behavior patterns in the e-Home system, a project aimed at providing assistive technologies for elderly individuals. The primary objective is to investigate the effectiveness of different prediction approaches, including hourly binning and clustering algorithms, for analyzing time-series data from infrared temperature sensors.

• Unsupervised prediction and clustering of user behavior patterns in the e-Home system.
• Analysis of time-series data from infrared temperature sensors.
• Comparison of different prediction approaches, including hourly binning, hierarchical clustering, and Dirichlet Process GMM.
• Evaluation of the effectiveness of these methods for predicting user behavior.
• Assessment of the impact of data quantity and distribution on prediction accuracy.

Zusammenfassung der Kapitel (Chapter Summaries)

The thesis begins with an introduction that outlines the motivation for this research, the goals of the project, and the methodology used. Chapter 2 provides a theoretical background on Gaussian Mixture Models, Dirichlet Process GMM, and hierarchical clustering, which are the key algorithms explored in the thesis.

Chapter 3 delves into the implementation details of the project, covering the software specification, data and sensors used, data visualization and inspection, sensor value discretization and extraction, unsupervised event extraction, data structure for event analysis, data quality and quantity, and predictive analysis.

The chapter on predictive analysis explores three different methods: hourly binning analysis, hierarchical clustering analysis, and Dirichlet Process Gaussian Mixture Model clustering analysis. Each method is explained in detail and evaluated on different data sets. The chapter also discusses the advantages and disadvantages of each approach.

Schlüsselwörter (Keywords)

The main keywords and focus topics of this thesis include assistive technologies, e-Home system, infrared temperature sensors, unsupervised prediction, clustering, hourly binning analysis, hierarchical clustering, Dirichlet Process GMM, time-series data, user behavior patterns, data quantity, and prediction accuracy.