完整後設資料紀錄
DC 欄位 | 值 | 語言 |
---|---|---|
dc.contributor.author | 柯坤安 | zh_TW |
dc.contributor.author | 洪士林 | zh_TW |
dc.contributor.author | Ko, Kun-An | en_US |
dc.contributor.author | Hung, Shih-Lin | en_US |
dc.date.accessioned | 2018-01-24T07:39:42Z | - |
dc.date.available | 2018-01-24T07:39:42Z | - |
dc.date.issued | 2016 | en_US |
dc.identifier.uri | http://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT079870511 | en_US |
dc.identifier.uri | http://hdl.handle.net/11536/140732 | - |
dc.description.abstract | 本研究利用政府機關公開、透明的公共工程決標金額之大數據資料挖礦分析其影響變數因子及敏感度分析,以2011年至2015年為樣本區間進行資料分析,本研究蒐集之總案例數為9316例。經由大數據分類整理其決標案例,再透過隨機抽樣方式,挑選出有效資料1295例作為本研究樣本;分別再以亂數抽樣方式,區分為--260例、550例及840例等三組模組做為類神經網路之學習訓練組。另以--87例與130例二組資料做為類神經網路的測試驗證組。研究將針對訓練案例組與測試組進行交叉組合驗證。研究中擬定八項影響變數因子,分別為專案類別、工程種類、工程屬性、履約期限、產業類別、投標家數、底價金額及預算金額。再藉由倒傳遞類神經網路學習演算法(BPN)與迴歸分析因子者間的重要性,探討預測工程決標金額與影響變數因子的關係。研究結果顯示,於BPN神經網路模型中驗證組之準確率(平均相對誤差百分比)優於總平均正確率。BPN之學習效果以260例網路架構為型6-1 <8-30-1>者為最佳。驗證成效則為測試組87例網路架構為型7-1 <8-35-1>者為最佳,其準確率為90.61%。至於迴歸分析模式,驗證成效最佳化模型為--迴歸模型二,預測相對誤差之變異值為5.87%。最終於大數據資料挖礦分析中87例,將影響預測決標金額的影響變數因子進行敏感度分析,其中以因子八--預算金額的敏感度係數值為最高。研究結果顯示,藉由BPN類神經網路之敏感度分析歸納出公共工程決標金額與預算金額有明顯的關聯性。 | zh_TW |
dc.description.abstract | This work presents a study of big data mining via artificial neural network models using the government agencies open and transparent public works bidding data. There are 9316 Taipei City public construction works bidding data from 2011 to 2015 were utilized as training and analyzing data. Among them, 1295 of data were selected randomly as study samples. Following, the sample data were random classed as five groups. Three are training groups with 260, 550, and 840 cases, respectively. Two are analysis groups with 87 and 130 cases. Eight influence factors are considered in this work. They are project category, engineering type, engineering attributes, duration of performance, industry category, bidder numbers, project reserve amount, and budget amount. The goal of the work is to analyze and identify the importance of influence factors and categorize sensitivity among these factors to the amount of the bid. Artificial neural networks with error backpropagation learning algorithm (BPN) and regression analysis are employed respectively to analyze between the predictive engineering award amount and the effect of variable factors. Research results show that learning effect of 260 cases of BPN-network architecture for type 6-1 <8-30-1> is the best. Validation results for test group of 87 cases of network architecture for type 7-1 <8-35-1> is the best, its accuracy is 90.61%. For regression analysis model, model two performed the best result, variation of the relative error of the predicted value is 5.87%. Finally, the group with 87 cases were employed in big data mining analysis to identify the sensitivity among influence factors. Analysis results revealed that the factor, budget-amount of the project, is the uppermost influence factor to the amount of the bid. Research shows that sensitivity analysis of neural network using BPN summed up public works bids amounts with budgeted amounts have obvious relevance. | en_US |
dc.language.iso | zh_TW | en_US |
dc.subject | 大數據 | zh_TW |
dc.subject | 資料挖礦 | zh_TW |
dc.subject | 決標金額 | zh_TW |
dc.subject | 底價金額 | zh_TW |
dc.subject | 預算金額 | zh_TW |
dc.subject | BPN倒傳遞類神經網路學習演算法 | zh_TW |
dc.subject | 迴歸分析 | zh_TW |
dc.subject | 敏感度分析 | zh_TW |
dc.subject | Big data | en_US |
dc.subject | data mining | en_US |
dc.subject | the amount of the bid | en_US |
dc.subject | the amount of the reserve price | en_US |
dc.subject | budget | en_US |
dc.subject | BPN back-propagation neural network | en_US |
dc.subject | regression analysis | en_US |
dc.subject | sensitivity analysis | en_US |
dc.title | 應用類神經網路於大數據資料挖礦分析研究—以台北市2011年至2015年工程採購決標金額之標案為例 | zh_TW |
dc.title | Application of neural network analysis of big data mining—Case study of engineering procurement tenders award amount of the Taipei City in 2011 to 2015 | en_US |
dc.type | Thesis | en_US |
dc.contributor.department | 工學院工程技術與管理學程 | zh_TW |
顯示於類別: | 畢業論文 |