Course syllabus - Data analysis, clustering and classification

Objectives

The goal of the course is to give knowledge of different methods for solving clustering and classification problems within data analysis. The course is intended to give both understanding of how the methods work as well as give training into how these methods can be used in different practical problems and applications. The course will present concrete examples of applications where clustering and classification appears and give examples of other aspects of data analysis, such as evaluation of results and common data transforms.

Learning outcomes

At the end of the course the student is expected to be able to:
 
1. describe different methods for classification and clustering of data as well as discuss some of their properties, advantages and disadvantages,
2. independently solve clustering and classification problems of given data.,
3. describe and apply different types of simple data transforms and methods for dimension reduction to problems in data analysis,
4. describe what is meant by the "Kernel trick", the mathematics behind the method and how it can be applied in some methods for clustering or classification,
5. evaluate the results of a clustering or classification,
6. give examples of applications where clustering and classification problems occur.

Course content

- Methods for clustering such as: k-means, Expectation maximization (EM), DBSCAN.
- Methods for classification such as: Naive Bayes, Support Vector Machines (SVM), Linear Discriminant Analysis (LDA).
- Methods for dimension reduction such as: Principal Component Analysis (PCA), Non-negative Matrix Factorization (NNMF).
- Hilbert spaces, positive definite functions, reproducing kernel Hilbert space.
- Common data transforms: z-transform, angles/time of day to Cartesian coordinates, "Kernel trick" applied in for example SVM and Kernel k-means.
- Methods for evaluation: crossvalidation, Precision, Recall, F-measure, Receiver Operating Characteristic (ROC)
- Discussion and examples of applications of clustering and classification of different types of data.

Specific requirements

Linear algebra, 7.5 credits or Applied matrix analysis, 7.5 credits or the equivalent and Probability Theory and Statistical Inference, 7.5 credits or the equivalent and Fundamentals of programming, 7.5 credits or the equivalent. In addition, Swedish B/Swedish 3 and English A/English 6 are required. In cases when the course is offered in English, the requirement for Swedish B/Swedish 3 is excluded.

Examination

LAB1, 1.5 credits, grades Fail (U) or Pass (G): computer based laborations concerning learning outcomes 1, 3, 5, 6.
PRO1, 3 credits, grades Fail (U), Pass (G) or Pass with distinction (VG): Independent solution of a given problem concerning learning outcomes 2, 3, 5
TEN1, 3 credits, grades Fail (U), Pass (G) or Pass with distinction (VG): exam concerning learning outcomes 1, 3, 4.

A student who has a certificate from MDU regarding a disability has the opportunity to submit a request for supportive measures during written examinations or other forms of examination, in accordance with the Rules and Regulations for Examinations at First-cycle and Second-cycle Level at Mälardalen University (2020/1655). It is the examiner who takes decisions on any supportive measures, based on what kind of certificate is issued, and in that case which measures are to be applied.

Suspicions of attempting to deceive in examinations (cheating) are reported to the Vice-Chancellor, in accordance with the Higher Education Ordinance, and are examined by the University’s Disciplinary Board. If the Disciplinary Board considers the student to be guilty of a disciplinary offence, the Board will take a decision on disciplinary action, which will be a warning or suspension.

Grade

Three-grade scale