2D5342 Knowledge Discovery and Data Mining
Graduate Course 4p
Proliferation of IT and automatic surveillance applications
means that vast volumes of data are gathered
for operations and other purposes. In business there is an
awareness that data collected constitute an untapped resource of
business knowledge, that can be used to commercial advantage. In
sciences, data collected in large scale experiments have rapidly
outgrown the communities resources to analyse data. The method of
using advanced 'semiintelligent' systems to systematically 'mine'
data repositories and create knowledge is known as
knowledge discovery and data mining.
In this course we introduce the main methods used in the area,
as supported by techniques originating in the Data Base, Artificial
Intelligence, Statistics and Visualization fields. We particularly
cover the theoretical and practical problems in identifying 'true and
knowledge as opposed to erroneous, random and uninteresting knowledge,
a problem much
studied in statistics and data base practice.
Graduate student standing(forskarstuderande) or undergraduate student(teknolog) with first
courses passed in programming, statistics, and (recommended) data bases
or information systems.
After passing this course, you will
- Know the fundamental approaches to knowledge discovery and data mining,
the main theoretical foundations, as well as its code of practice.
- Know about several tools in the area and be able to use at least one.
- Be able to follow research and development in the area
- Be able to assess the applicability of the technology for a particular
scientific problem area.
- KDD philosophy
- Bayes rule and its interpretation as inference tool.
- Data Cleaning, de-trending, etc
- Learnability, VC-dimension
- Statistical Techniques: MV analysis, SVD technique, etc
- Classification and clustering
- Bayesian networks and graphical models
- Prediction and sequence mining.
- Soft Computing Techniques: Fuzzy, rough and Neural computing
- Visualization techniques : Visualizers and VR techniques
- Markov Chain Monte Carlo methods
Examination is individual, and can consist of discussion of texts with me,
presentation in class, homeworks and/or a small project.
A list of papers you read for the course, preferrably with comments, and
proposals for course improvement should be turned in.
Ingång till Nadas kursweb