两阶段混合引导的偏好多目标优化算法

在偏好多目标优化问题求解中，基于偏好点引导方式的优化算法性能易受偏好点具体位置的影响，且不能控制偏好解集大小；而基于偏好区域引导方式的优化算法虽然能控制偏好解集规模，但算法初期收敛效率不够高。针对此问题，提出一种两阶段混合引导的偏好多目标优化算法。算法初期采取偏好点的引导方式，同时引入偏好界限动态调整策略，使得种群快速收敛到偏好区域附近。迭代一定次数后，采用偏好向量引导方式指导算法搜索，达到控制偏好解集范围的目的。与几种经典的偏好优化算法进行实验对比分析，结果表明，所提算法性能不受偏好点位置影响，得到的偏好解集能很好地表征决策者的偏好信息，并且控制了偏好解集范围，便于决策者的最终决策。

Preferential multi-objective optimization algorithm based on two-stage hybrid guidance

Solving the preference multi-objective optimization problem, the performance of the optimization algorithm based on the preference point guidance method is easily affected by specific position of the preference point and cannot control the range of preference solution set. While the optimization algorithm based on the preference region guidance method can control the range of preference solution set, but its initial convergence efficiency is not high enough. In response to the above problems, this paper proposes a two-stage hybrid guided preference multi-objective optimization algorithm. The algorithm adopts a preference point guidance method at the beginning of the algorithm, and introduces a dynamic adjustment strategy of preference boundaries, so that the population quickly converges to the vicinity of the preference area. After a certain number of iterations, the preference vector guide method is used to guide the algorithm search to achieve the purpose of controlling the range of the preference solution set. Finally, through experimental comparison and analysis with several classic preference optimization algorithms, the result shows that the performance of the proposed algorithm is not affected by the position of the preference point, and the preference solution set obtained by the proposed algorithm can well represent the preference information of decision makers and control the range of preference solution set, which is convenient for the final decision of the decision maker.