Data mining

Course goals

After this course the student

• knows how important data-mining algorithms work, and what their properties are
• knows how to interpret the models and patterns produced by these data-mining algorithms
• knows how to match a data-analysis problem with the appropriate data mining algorithm(s)
• understands general concepts of data analysis, such as overfitting, the curse of dimensionality, the bias-variance decomposition and the VC-dimension
• knows how to perform a (comparative) data-mining experiment in a sound manner
• has gained experience with data mining algorithms through one or more assignments.

Assessment
The course is graded through:

• a digital exam
• one or more assignments
• homework exercises

Prerequisites
It is assumed that the participant has knowledge of:

• algorithms and data structures.
• probability theory and statistics.
• calculus.
• linear algebra.

This is a required course for students in the Data Science (DASC) master program.

Content

Topics covered may include (content can vary somewhat from year to year):

• PAC learning and the VC-dimension
• the bias-variance decomposition
• classification tree algorithms
• bagging, (gradient) boosting and random forests
• graphical models (including Bayesian networks)
• frequent pattern mining (sets, sequences, trees)
• logistic regression
• text classification (with the multinomial naive Bayes model)
• classification for network link prediction
• Support Vector Machines

For further details see the course's website https://ics-websites.science.uu.nl/docs/vakken/mdm/

Course form
Lectures, tutorial sessions.

Literature
Selected book chapters, articles, and lecture notes.