差分演化演算法( Differential Evolution;DE )是近年來演化式計算的熱門演算法之一,它擁有隨機搜尋的方法且具有優越的效能,因此常被應用於資管領域之中,如:資料探勘、排程、路徑規劃、決策支援等,但差分演算法也存在演化式計算的缺點,如容易陷入區域最佳解、收斂不穩定等。 本研究提出以差分演算法和粒子群最佳化演算法( Particle Swarm Optimization;PSO )為基礎的DEPSO演算法(Differential Evolution Particle Swarm Optimization;DEPSO )進行改良。DEPSO透過雙演化策略(Dual Evolution Strategy;EDS)及資訊分享的師徒式機制,結合兩演算法之優點並互補彼此之缺點,使演算法在運算過程中,減少迭代的運算次數得到最佳解,並改善差分演算法的缺點,透過實驗結果證明DEPSO的確有效改良差分演算法之求解成效及在收斂上的穩定性。
Differential Evolution (DE) is one of the novel algorithms of evolution computation. Although it performs superiorly, DE has several disadvantages. In this study, we proposed the construction of a novel DEPSO algorithm in DE and Particle Swarm Optimization (PSO). DEPSO is a strategy of Dual Evolution (DES) based on the master-apprentice mechanism for sharing information. During the iteration, between the two algorithms can be iterative operation to improve the drawbacks “easy to drop into region optimum” moreover increasing the performance to obtain the advantage of accuracy solving and stable convergence.