When an experimenter performs experiments, the most important consideration is the experimental design. However, if the experimental cost is low, the experimental design is not a problem because the experimenter is at liberty to choose any design he wants. But if the cost is high and the number of experiments is restricted, then experimental design becomes critical.
Classical experimental designs are the best designs and they have unique properties. But they can not be used when the experimenter is confronted with special problem such as a restriction on the number of experiments, restricted experimental runs, augmentation of designs, or special shapes of an experimental region. In such cases, better designs can be obtained by using a criterion yielding optimal designs.
Until now, algorithms for constructing exact D-optimal designs assumes that costs for running experimental points are the same. However, this assumption is frequently violated in practice.
The purpose of this thesis is to develop a new heuristic algorithm for constructing exact D-optimal designs when costs for running experimental points may be different and the budget constraint is given. Computational experiments are also conducted to evaluate the performance of the proposed algorithm.
