Progressively interactive evolutionary multiobjective optimization

Thumbnail Image

Files

Aalto_DD_2011_017.pdf (2.9 MB)

URL

Journal Title

Journal ISSN

Volume Title

School of Business | Doctoral thesis (article-based) | Defence date: 2011-03-16
Checking the digitized thesis and permission for publishing
Instructions for the author
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.

Authors

Date

2011

Department

Liiketoiminnan teknologian laitos

Major/Subject

Quantitative Methods
Taloustieteiden kvantitatiiviset menetelmät

Mcode

Degree programme

Language

en

Pages

vi, 129 s.

Series

Aalto University publication series. DOCTORAL DISSERTATIONS, 17/2011

Abstract

A complete optimization procedure for a multi-objective problem essentially comprises of search and decision making. Depending upon how the search and decision making task is integrated, algorithms can be classified into various categories. Following `a decision making after search' approach, which is common with evolutionary multi-objective optimization algorithms, requires to produce all the possible alternatives before a decision can be taken. This, with the intricacies involved in producing the entire Pareto-front, is not a wise approach for high objective problems. Rather, for such kind of problems, the most preferred point on the front should be the target. In this study we propose and evaluate algorithms where search and decision making tasks work in tandem and the most preferred solution is the outcome. For the two tasks to work simultaneously, an interaction of the decision maker with the algorithm is necessary, therefore, preference information from the decision maker is accepted periodically by the algorithm and progress towards the most preferred point is made. Two different progressively interactive procedures have been suggested in the dissertation which can be integrated with any existing evolutionary multi-objective optimization algorithm to improve its effectiveness in handling high objective problems by making it capable to accept preference information at the intermediate steps of the algorithm. A number of high objective un-constrained as well as constrained problems have been successfully solved using the procedures. One of the less explored and difficult domains, i.e., bilevel multiobjective optimization has also been targeted and a solution methodology has been proposed. Initially, the bilevel multi-objective optimization problem has been solved by developing a hybrid bilevel evolutionary multi-objective optimization algorithm. Thereafter, the progressively interactive procedure has been incorporated in the algorithm leading to an increased accuracy and savings in computational cost. The efficacy of using a progressively interactive approach for solving difficult multi-objective problems has, therefore, further been justified

Description

Supervising professor

Korhonen, Pekka, professor

Keywords

Other note

Citation

Permanent link to this item

https://urn.fi/URN:ISBN:978-952-60-4052-3

Collections

[diss] Kauppakorkeakoulu / BIZ