基因演算法廣泛地被運用於各種不同問題的最佳化排程上，且都證明有不錯的求解品質。然而基因演算法本身並沒有處理限制條件的能力，但在實際工程問題上，大部分為包含限制條件的最佳化問題。因此必須以額外的機制將限制條件問題，轉換為無限制條件問題。一般在基因演算法中用來處理限制條件的方法有四種：1.保留合理解，非合理區的解完全刪除。2.採用懲罰函數法。3.採取分開合理區和非合理區的解來處理。4.混合法。 一般結合基因演算法與限制條件滿足問題，在基因運算子演算過程中，並沒有加入限制條件的概念。為分段式求解過程，先經由基因運算子產生新個體後，再判斷是否符合限制條件。如不符合則再重新產生新個體，以試誤法的方式直到產生符合限制條件為止。因此搜尋點會落於不合理解區域，造成搜尋時間上的浪費。 本研究重新設計基因演算法的運算子及表示法，提出一限制型基因演算法(Constrained_GA)，將限制條件以基原的特性加入染色體的訊息中，使限制條件嵌入基因運算子演算過程。因此染色體中較佳的基原得以被保存，且將搜尋點落在部分合理解範圍內。本方法證明運用於線性最佳化問題及JSP(Job-Shop scheduling problems)問題上，可獲得不錯的求解品質。 本研究以基因演算法運用於醫療臨床路徑排程問題，初步探討基因演算法於醫療服務業的可行性。 The Genetic Algorithm (GA) has been widely applied in variant optimization problems and reported as satisfied in many cases. It is well known that the constraint features are everywhere in the real problem. Many research have been tried to include the constraints handling capability in the GA process. Usually, the constraints handling mechanism (CSM) has been integrated into the GA process as an add-on. After each iteration, the possible solutions will be checked via the CSM before be marked as the member of the new population. In this approach, many trial and errors will be conducted to display the constraints handling feature. However, in our approach, we integrate the constraint handling into the GA process. That is, for the new generated population, without repeated check and search, all members of the population will be guaranteed to satisfy at least one constraint. This approach has been successfully applied to the linear optimization problems and Job-Shop Problem problems respectively.
