應用複迴歸分析與倒傳遞類神經網路模式探討排水溝工程之相依性–以南投地區排水工程為例

An Application of Multiple Regression and Backpropagation Network for the Interdependence Model of Drainage-engineering-A Case study of Nantou Area，Taiwan.

呂騏竹(Chi-Ju Liu)

指導教授 ： 翁祖炘

國立臺北科技大學/工程學院/材料及資源工程系所/碩士(2006年)

摘要

工程成本預測及推估一直是工程界中最具重要且有高風險的作業，一旦初步判斷錯誤會影響整體工程之結果，對排水溝工程的相關單位而言，如果能夠更準確的預測工程費用可以減少採購發包、編列預算、審核底價時之時間花費及因估價造成之成本損失外，並可迅速執行工程內容，有利於對工程施工成本之效益執行查核與評估，且在有缺失時儘早研擬配套的改善措施。
筆者蒐集台灣省南投縣市地區於民國八十九年至九十二年間之排水改善工程之預算書，並參考專家們的意見及使用多變量統計分析，包括集群分析、單一因子多變量變異數分析、複迴歸分析與應用人工智慧方法之倒傳遞類神經網路等數理分析方法，預期藉由所得到的結果來評估複迴歸分析模式及倒傳遞類神經網路模式之優劣，並建立適用於南投地區之排水溝工程成本，以提供給相關單位控制工程造價，降低工程成本之參考依據。
由研究結果顯示：一、使用複迴歸及倒傳遞網路進行非線性推估效果皆佳，因此排水溝工程之發包經費呈現非線性分佈且其自變數與依變數之相關係數皆為高度相關。二、比較複迴歸及倒傳遞網路模式推估優劣得知，在預測績效的評估使用MSE以倒傳遞網路為佳，而使用MAPE皆呈現高精確度，故本研究具有可行性且有參考之價值。

關鍵字

集群分析 ； 單一因子多變量變異數分析 ； 複迴歸分析 ； 倒傳遞類神經網路 ； 排水溝之成本估價

並列摘要

Project cost prediction or estimate has been an important,yet highly risky process.A mistakenly preliminary judgment will affect a whole project outcome.As far as drainage project is concerned,if given an exact prediction or estimate, the institution concerned may greatly cut down costs including purchasing,contact process,budgeting,bidding, and some other costs and losses incurring from estimate. In addition,the exact prediction or estimate can facilitate the project,beneficial for overall project costs and efficiency in checking and estimate.In this way,we may bring up a set improvement measure at the earliest day to make up for shortage.
I personally gathered the budge plans for drainage improvement projects in the area of Nantou County,Taiwan during the period between 2000 to 2004,referred to specialists suggestions and applied multi-variables statistical analyses,including cluster analysis, one-way multivariate analysis of variance,multiple regression analysis and applying back-propagation network and other methods of mathematical analysis.It is expected the advantages and disadvantages of multiple regression analysis modes and back-propagation network modes are appraised by the acquired outcome,and applicable drainage project costs in Nantou area is then established to serve as reference for the relevant institutions,which may control over project costs and reduce project costs.
The findings of the study show 1. multiple regression and back-propagation network is applied to conduct non-linear estimation,which proves good performance.The contracting expenditures of drainage project thus displays non linear distribution and its variables are higly correlated with the coefficients of variables.2. It is inferred from the comparison of Multiple Regression and Back-Propagation Network that MSE is applied with back-propagation network is better in terms of the performance estimation,and the application of MAPE shows high precision.This study consequently is feasible and good in reference.

並列關鍵字

cluster analysis ； one-way multivariate analysis of variance ； analysis of multiple regression ； backpropagation ； cost estimation of drainage-engineering

參考文獻

[20] Prybutok,V.R,J.Yi and D.Mitchell, "Comparison of Neural Network Models With ARIMA and Regression for Prediction of Houston’s Daily Maximum Ozone Concentrations" ,European Journal of Operational Research, vol.122, 2000, pp. 31-40.

[21] Chang,N.B, Chen,Y.L and Chen,H.W, "A fuzzy regression analysis for the construction cost estimation of wastewater treatment plants﹙Ⅰ﹚Theoretical development",Journal of Environmental Science and Health,A32﹙4﹚,1997,pp.885-899.

[22] Smith,A.E and Mason ,A.K , "Cost estimation predictive modeling：regression versus neural network", The Engineering Economist ,vol.42, no.2,1997,pp.137-161.

[24] Cammartat,G..S,Gavalieri and A ,Fichera , "A Neural Network Architecture for Noise Prediction",Neural Networks,vol.8,no.6 ,1995,pp.963-973.

[25] Martin,C.A.&Witt,S.F, "Accuracy of econometric forecasts of tourism",Annals of Tourism Research,vol.16,1989,pp.407-428.