Introduction to economic modelling with computational intelligence (CI)

Sept. 5-9, 2016

Helsinki

Estimated workload

10 h independent work prior to course material at: http://www.mathworks.se/help/pdf_doc/fuzzy/fuzzy.pdf

115 h Model construction and report writing.

22 h of lectures

ECTS

5

Venue & Time table

University of Helsinki, Viikki Campus, computer class 155A in building A. 

Sept. 5-7 at 10-14
Sept. 8 at 12-16
Sept. 9 at 9-13

Also possible for everyone to follow the course live elsewhere by using the Adobe Connect distant education service. 

Link to Adobe Connect

Course description

(log in as guest, no specific username required, you can give e.g. your name when logging in)

The course aims at providing basic knowledge and skills for constructing computer models with computational intelligence methods. Skills for economical applications in this area are also provided. To some extent, knowledge and skills of statistical modelling are empowered. A cutting-edge, multi-disciplinary approach to model construction will be adopted.

Computational intelligence (CI) generally includes such methods as neural networks, fuzzy systems, evolutionary computing, Bayesian networks, cellular automata and swarm theory. Today CI is applied to systems and models in control, decision making, pattern recognition, robotics, data mining, biological and social modeling and statistics, inter alia. In particular, they are useful when non-parametric or non-linear models are constructed, and they often yield better and simpler computer models than the corresponding classical mathematical models. 

Many models and theoretical results of CI are already available in economics, but their application is still uncommon and fairly unfamiliar in the Nordic countries (possibly excluding Finland). Hence, there is a justified need for a course which deals with the basics and economical applications of CI. 

The course focuses on fuzzy systems, neural networks and evolutionary computing. It presents the basic principles and economical applications of these methods in computer modeling. Corresponding traditional mathematical and statistical models are also considered. During the course days fuzzy rule-based economical computer models are designed and constructed, and they are fine-tuned with neural networks and genetic algorithms. Fuzzy cognitive maps, both numerical and linguistic, are also examined because these are simple and usable models for examining complex phenomena. CI models are compared to such corresponding traditional methods as regression, cluster, discriminant and time series analysis. 

Matlab software is used for the assignments, but previous knowledge of the software is not required. Basic (high school level) knowledge of mathematics is recommended. The course days include 4-5 hours of lectures, which can also be followed live with Adobe Acrobat connection. Lectures comprise of theory parts on computational intelligence and exercises on economical model constructions with Matlab. A course report should be completed and sent for evaluation within a month after the on-site course days. The report (5-7 pages) should present a simple economical application of fuzzy, neuro-fuzzy or genetic-fuzzy model. These application ideas will be considered and discussed with the participants during the course. 

Moodle learning environment is used for learning materials, discussions and for delivering and evaluating the course reports.

Learning outcomes

The students shall be given possibilities to work with fuzzy modelling in economics. The students will see an introduction to newly developed methods in neural networks and genetic algorithms. 

Knowledge: highly specialised, novel, cutting-edge and multi-disciplinary knowledge on computational intelligence modelling, especially in economic applications. 

Skills: specialised problem-solving skills for constructing computer models with computational intelligence. 

Competence: ability to manage and transform computational intelligence modeling to work or study contexts that are complex, unpredictable and require novel strategic approaches.

Nordic dimension - reasons why the course should be facilitated through NOVA

Similar courses are not provided in any of the Nordic countries and there's a clear need for such know-how.