基于次大值标杆的秘书问题研究

秘书问题的实质是决定何时停止观察选项、而不是哪一个选项被选择,已有解决秘书问题的策略,其主要特征是以取样选项中的一个最大值作为标杆。该策略的优点是能保证命中概率最大,然而其不足是很少考虑决策者的有限理性与启发式偏见,因此本文提出了次大值标杆的设想,然后从理论上计算出该策略的最优截止阀值与命中概率,并通过计算机仿真实验验证与比较了该策略的特征与规律。研究结果发现在最大化命中概率的条件下,标杆降低导致取样观察选项的数量不断增加,但命中概率却逐渐降低。

Secretary Problem Study Based On the Sub-maximum Benchmark

The essential decision in "secretary problem"(SP) is when to stop searching additional information and not which alternative to be chosen.An established optimal policy is characteriged by the strong point of the best winning probability,but it does not take into account the bounded rational of decision-maker.We propose the alternative sub-maximum benchmark,and find that the decreasing benchmark corresponds to the increasing magnitude of the sampled applicants on the ground of the maximized winning probability.However,we also discover the decreasing trend of the winning probability with the decreasing benchmark.