Data Fusion and Analysis

Recommended Prerequisites

Linear algebra, probability theory, Signal Processing, MATLAB

Teaching Methods

The teaching/learning process follows a learning approach of collaborative type, based on the presentation, analysis and discussion of concepts and techniques.
Theoretical classes (2 hours/week): Presentation of the concepts, principles and fundamental techniques for information fusion.  Examples of concrete situations will be presented, to illustrate the practical interest of the techniques and its application to real conditions.
Practical classes (2 hours/week): Practical problems addressing the theoretical concepts, their analysis, discussion and implementation.

Learning Outcomes

This course (Information Fusion) aims to introduce fundamental concepts related to the theory, the design and the implementation of data/information fusion from various sources in various domains. The course covers distinct methodologies capable of combining data from multiple, heterogeneous sources, in order to make inferences with application in decision support systems, clinical diagnosis, fault detection and others. The content is multidisciplinary, including topics such as time series analysis, state estimation, stratification, classification and data mining.
This course will enable students to understand, identify, select and implement distinct information fusion techniques. In particular,students should after the course have the ability to characatrize the diferent architetures, methods and algorithms for information fusion, and to be able to apply these to pratical situations in distinct domains.

Work Placement(s)

No

Syllabus

Introduction and concepts
Definitions, benefits, challenging problems and applications of information fusion
Types of data fusion, data fusion models, data fusion architectures, data fusion Levels

Modules
1. Data-related fusion aspects
Errors in raw data: inconsistency, imperfection, missing data, noisy data, outliers.
Data preparation: alignment, transformations.

2. Data fusion algorithms
Data level approaches: weighted average, Kalman and Extended Kalman Filter, Particle filtering.
Feature level approaches - Parametric and Non-Parametric: K-Nearest neighbour, Decision Tree, Neural Networks, Support vector machines, Gaussian mixture model, k-Means.
Decision level approaches: Classical inference, Bayesian inference, Dempster-Shafer Theory, Markov chain Monte Carlo, Fuzzy logic.
Ensemble methods: voting, adaboost, random forest.

3.Comparison and Performance of fusion systems
Accuracy and dimensionality.

4. Applications
Biomedical applications, Fault detection, Visualization, Networ

Head Lecturer(s)

Jorge Manuel Oliveira Henriques

Assessment Methods

Assessment
Project: 50.0%
Exam: 50.0%

Bibliography

Data Fusion: Methods, Applications and Research; V. Albert, E. Aba; 2017 Research Methodology and Data Analysis; ISBN: 978-1-53612-720-1.

Information Fusion and Analytics for Big Data and IoT; A. Bossae, B. Solaiman; 2016, Artech House; ISBN-13: 978-1630810870.

Concepts, Models, and Tools for Information Fusion; E. Bosse, J. Roy, S. Wark; 2007, Artech House, ISBN 1596930810 (ISBN13: 9781596930810)

Data Fusion: Concepts and Ideas; H.B. Mitchell; 2012 Springer; ISBN 978-3-642-27221-9; DOI DOI 10.1007/978-3-642-27222-6

Sensor and data fusion : a tool for information assessment and decision making; L. Klein; 2010 SPIE; ISBN 0-8194-5435-4.