A New Selection Procedure For Large Scale Problems

Loading...
Thumbnail Image
Date
2012-04
Authors
Almomani, Mohammad Hani
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
This thesis considers the problem of selecting the stochastic system that has the best (maximum or minimum) expected performance measure when the number of alternatives is finite but large. Ranking and selection procedures have been used successfully for solving problems with small number of alternatives. In order to reduce the computational problem, the idea of ordinal optimization is being used with the objective of finding a good enough system instead of looking for the best system. In this thesis, a new selection approach is proposed for selecting a good system when the number of alternatives is very large. This approach contains four stages; In the first stage, the ordinal optimization procedure is used for selecting a small subset that overlaps with the set that contains the actual best m% systems with high probability. Then, the optimal computing budget allocation procedure is applied in the second stage to allocate the available computing budget. This is followed with the subset selection procedure to get a smaller subset that contains the best system among the subset that was selected before with high probability. Finally, indifference-zone procedure is used to select the best system from the previous subset with high probability. The efficiency of the proposed selection approach is being examined from two different points. First, based on some parameters changing such as the initial sample size, increment in simulation samples, total budget, and the elapsed time. Secondly, based on three sets of the stopping rules such as sequential, expected opportunity cost and probability of good selection of the stopping rule. In addition, comparisons between the proposed selection approach and the Three-stage selection approach are also presented. Finally, one of the most difficult problem in designing of production lines, which is known as buffer allocation problem is presented as a real application for the proposed approach. The implementations of our approach are presented with some numerical examples. The results show that in general, the proposed selection approach made the correct selection with high probability.
Description
Keywords
New selection approach is proposed for selecting , when the number of alternatives is very large
Citation