| 摘要: | 電力之於人們的日常生活日趨重要。為因應此需求，加快電力輸配電系統的發展刻不容緩，主要是特高壓（EHV）輸電鐵塔，因為特高壓鐵塔用於傳輸全臺灣的電力。為了建造此類鐵塔，須由台灣電力公司（TPC）招標，以便EPC公司可以執行施工。特高壓輸電鐵塔的建設主要分為兩個部分：建設特高壓鐵塔基礎並建設連接特高壓鐵塔。為了確定投標價格，台電對每個項目進行初步分析，以找到投標的最佳預測價格。然而，這個過程需要很長時間，並且根據塔的數量投注大量的人力。因此，本研究旨在確定預測特高壓鐵塔基礎項目成本所需的關鍵因數，從而減少準確預測上述成本所需之時間與人力。此外，本研究並研究變量的聚類特徵，以發現這些因數之間有趣的關係。為了達到這個結果，k-means clustering用於根據成本對數據集中的每個關鍵因數進行聚類。此外，基於規則的相關性同時採用Apriori 演算法尋找關鍵因數之間的有趣關係。分析發現，從找到的十三個關鍵變量中，單基樁（X2）、人工挖方（X5）與普通模板（X7）對特高壓鐵塔基礎造價的貢獻較大，表現在其他集群中成本最高的第二個集群中這三個因數的數量很大。對於基於規則的關聯，一共找到了1174條法則，其中選擇了102條法則，有2個前件和1個後件。從這 102 條法則中，有 62 條法則的可靠度為 1.000，從這 62 條規則中，找到了 6 條提升度最高的法則。所有102條規則的提升度皆高於1，這意味著每條規則都有其意義，6條最高提升度的規則為9.2，這些規則雖然支持度較低但可靠度高，這 6 條規則是其中102 條規則中最重要的。;The needs of electricity in our daily lives is becoming more and more essential. To accommodate such needs, the development of electrical transmission and distribution system requires to be sped up, mainly on the Extra High Voltage (EHV) Transmission Tower due to the importance of the tower to transmit power throughout the country. In order to construct such tower, Taiwan Power Company (TPC) has a possibility to announce a tender so that an EPC company can carry out the task. The construction of EHV Transmission Tower is divided into two categories: one is to construct the EHV Tower Foundation, and the other is to construct the EHV Tower and the connection. To determine the tender price, TPC has been doing initial analysis for each project to find the optimum forecast price for the tender. However, this process takes a long time and depending on the number of towers, this probably takes a large amount of manpower as well. Therefore, this study aims to determine the critical variables needed to forecast the cost for EHV Tower Foundation project, so that it can reduce the time and manpower needed to accurately predict the said cost. Other than that, this study also investigates the variable’s cluster characteristics, and discovers interesting relationship between the selected variables. To achieve the result, k-means clustering is applied to cluster each variable from the dataset based on the cost. Aside from that, rule-based association is also employed together with Apriori algorithm to find the interesting relationship between the selected variables. From the analysis, it is found that single pile foundation (X2), manual excavation (X5), and common formwork (X7) contribute greatly to the cost of the EHV Tower Foundation from the 13 selected variables found. This is also shown by the high amount of these three variables within the 2nd cluster which has the highest cost among the other clusters. As for rule-based association, with minimum support of 0.1, a total of 1174 rules are found, where 102 rules with 2 antecedents and 1 consequent are selected. From these 102 rules, 62 rules have 1.000 confidence, and from these 62 rules, 6 rules with the highest lift are found. All of the 102 rules have lift value more than one, which means every rule has its own significance with the highest value of lift of 9.2 which has lower number of support, but high confidence, making these 6 rules the most significant out of the other 96 rules. |
