企業在進行決策規劃時,會先由高階決策者訂定整體的方向和目標,再將任務細分給次階的決策人員施行並達成次目標,這種階層式的決策問題可以經由多階的數學規劃來進行模擬。其中高階決策者和次階決策者各自控自不同的變因,卻相互受到牽制,而高階決策者對變因的控制也可視為對次階目標的一種試探性行為。 由於粒子群最佳化演算法具有模仿生物群體依賴相似特性之群體智慧(Swarm Intelligence)的概念方法,及粒子經驗交換及傳承世代之演算模式,其利用粒子族群具有探測(Exploitation)與開發(Exploration)的特色,可用於問題空間中搜尋全域的最佳解。本研究將運用粒子群最佳化演算快速收斂的特性來找出最低成本解,並發展出適用於求解線性和線性二階規劃的改良式粒子群最佳化,並將演算結果和過去使用基因演算法進行討論分析。結果顯示傳統粒子群和本研究所改良的粒子群演算法,再求解線性二階規劃的能力上皆有優異的表現,而其中又以改良式粒子群演算法的求解結果較為突出。
In decisions making for an organization, the upper-level decision maker has to determine the operation direction and goals first, and then forward them to the subordinate level as the base of decision making. Basically, subordinate-level manager has to achieve his/her goal without conflicting to high-level decision. This kind of hierarchical characteristics can be modeled and programmed by using mathematical programming. Particle Swarm Optimization (PSO) can mimic cooperation between individuals in the same group by using swarm intelligence and exchange experiences from generation to generation. To exploit and explore the hyperspace global optimal with PSO has many advantages, especially converges fast. This research attempts to develop a noval PSO named vector-controlled particle swarm optimization (VCPSO) to solve bi-level programming more accurately with comparison to genetic algorithm (GA) . The experimental results show that the proposed VCPSO is able to converge faster and has better accuracy than conventional PSO and GA.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。