如期、如質、如度完成工程專案,為工程專案管理之三個重要領域,儘管學理上發展出許多排程模式如甘特圖、平衡線圖、網狀圖或模擬,然而工期估算不如成本,有一快速簡單又準確量化之評估方法,亦不如品質有明確之規範或標準可資遵循,致使業主往往依賴工程師經驗概估工期,或依政策目標或營運目標而訂定工期,而廠商又限於備標時間的短暫,亦無能力正確估算工期,工期爭議問題日益頻繁。 本研究目的係透過倒傳遞類神經網路建構工期推估模式,以期提供主辦機關較準確且快速地進行工期推算,亦可提供廠商投標之際,匆促的時間壓力下能夠快速估算工期,俾使提供設計單位與廠商或主辦機關能夠快速且準確地做工期推估,並可作為合約工期訂定之參考。 To get projects done on time, within budget and at a certain quality level has been the main concerns of a construction project. However unlike budget and quality, there is neither a systematic way to estimate duration of a project nor standard specifications to be followed. Therefore this research aims at establishing a duration estimating model by analyzing the WBS and identifying the duration of different categories of architecture construction projects with application of expert interviews. Summarizing the duration influencing factors, it then applies Back-Propagation Network technique with practical cases to implement the model.