應用類神經網路於高科技廠專案工程成本估算之研究

Abstract

由於建置化學品相關系統之專業程度高，大多數高科技廠皆採用標案競標方式，而第三方專業工程公司對化學品相關系統建置之專案成本的估算，對其一個工程專案成功與否是重要之因素。現階段國內化學品相關系統之承包商之工程成本估算，大都僅憑專業估算人員，根據規模相似之工程經驗資料，缺乏一套系統化之工程成本估算方法。本研究以某一工程設備承包廠商在2003年至2012年間，共計79筆實際標案資料為案例，先透過線性迴歸方程式挑選影響專案成本之顯著變數其次，再利用類神經網路估算專案成本。由案例驗證可知，類神經網路可有效地地預估專案成本，其預估平均準確率約為97.22%。故可先透過本研究所建置的類神經網路預估模型估算專案成本，作為後續精算的參考依據。

The chemical related systems of the most high-tech factories were built by the third-party engineering construction companies via the competitive bidding. In order to acquire the project, the engineering construction companies have to accurately estimate the project cost. Traditionally, most domestic companies of the engineering construction in Taiwan usually estimate the project cost of the chemical related systems based on the professional experience of employees. In fact, the project cost could not be easily estimated because the employees might not have enohgh experience of the project cost estimation. In this paper, the real project data of an engineering construction company from 2003 to 2012 were taken for 79 real cases. Firstly, the stepwise multiple liner regression was used to identify the significant variables. After that, the significant variables were regarded as input data, the estimation model of project cost was established by using the artificial neural networks. According the results, the accuracy rate of the project cost estimation is about 97.22%. In other words, the artificial neural networks can generate the reasonable project cost. Therefore, they can accurately estimate the project cost via the estimation model established by the artificial neural networks.