拜現代科技發產迅速所賜，資料的存取越來越方便，這大大的降低了使用機器學期的門檻。比起傳統以經驗定勝負的方式，現在有更多的資訊可提供決策者做選擇，然而在仰賴機器學習進行決策時，資料的解讀方式或不同是模型的選擇可能會讓提供的資訊有所差異。因此，集成學習內的模型融合技術，能將各方模型所做的決策機制做融合，以達到「三個臭皮匠，勝過一個諸葛亮」的目的。也提供決策者一個更可靠、低風險的決策建議。
本研究將透過模擬軟體建立自來水供水管網之模擬資料，以流量、管徑等資料建立兩個獨立模型，用以作為管段問題偵測使用。後再加上集成學習的模型融合技術，將兩模型做合併，以達到合併、優化兩模型預測能力的目的。

Technology grows fast these days, which makes it easier to access tons of data for machine learning. It becomes more convenient to get a bunch of information to help decision-makers, rather than decide everything by pure experiences. However, machine learning can be a Double-side blade. The method and angle you choose to analyze data matters a lot, it may help decision-makers do the right thing or take them into the wrong way. Fortunately, there is a model fusion technique in the Ensemble learning field. It can help us combine several models into a single model and take potential advantage from every one of them. That can provide a better suggestion to decision-makers.
This research first used Epanet software to simulate the pipeline data and then two independent models were built to predict pipeline problems. It is observed that the combination of both the Model fusion technique to give a better prediction for problem detection.

[1] 陳明洲，2019，「台北自來水管網水理分析應用之規劃發展」，中華民國自來水協會
[2] Springer Science & Business Media, 2008, Fluid Mechanics: An introduction to the theory of sluid flows, p221
[3] Rossman, L.A., EPANET 2: users manual. 2000.
[4] 鄭明淵、邱建國、邱永芳、吳育偉、徐梓隆、宏興、林正軒，2014，「中國土木水利工程學刊」，p293-308
[5] Min-Yuan Cheng, Doddy Prayogo，2015. Symbiotic Organisms Search: A new metaheuristic optimization algorithm. Dept. of Civil and Construction Engineering, National Taiwan University of Science and Technology
[6]Jain, A. K., Duin, R. P. W. and Mao, J. (2000), Statistical pattern recognition: A review, IEEE Transactions on Pattern Analysis and Machine Intelligence 22(1), 4–37.
[7]Witten, I. H. and Frank, E. (2005), Data Mining: Practical Machine Learning Tools and Techniques, The Morgan Kaufmann Series in Data Management Systems, second edition, Morgan Kaufmann, San Francisco, CA.
[8]Bishop, C. M. (2006), Pattern Recognition and Machine Learning, Information Science and Statistics, Springer, New York.
[9]Aboozar Taherkhani, 2019, Adaboost-CNN: An adaptive boosting algorithm for convolutional neural networks to classify multi-class imbalanced datasets using transfer learning.
[10]J. Llinas, D.L. Hall, 1998, An introduction to multi-sensor data fusion
[11]許正憲，2008，資料融合方法應用於資訊電子產業績效評估模式之探討
[12]林君娥，2007，資料融合技術應用於事故影響下高速公路旅行時間預測之研究
[13]D.M. Mohan, 1986, Modified Hazen-Williams formula
[14] 陳泰旺、王美期，2011，「以EPANET建構水力分析模型，作為管線抽換及新設延長管線分析依據─以斗六系統為例」，經濟部台灣自來水股份有限公司
[15] P.L.Palmer, J. Kittler, M. Petrou, 1994, Using focus of attention with the hough transform for a accurate line parameter estimation
[16] Xiuyang Li, 2016, Improved Adaboost algorithm and object recongnition based on haar-like training
[17]Hence, V., 6.1 Energy and Momentum Coefficients
[18] Colebrook, C.F., et al., CORRESPONDENCE. TURBULENT FLOW IN PIPES, WITH PARTICULAR REFERENCE TO THE TRANSITION REGION BETWEEN THE SMOOTH AND ROUGH PIPE LAWS.(INCLUDES PLATES). Journal of the Institution of Civil engineers, 1939. 12(8): p. 393-422.