臺灣博碩士論文加值系統

近年來節能減碳成為各國家發展的主體，為了降低排碳量和解決交通壅塞問題，公共自行車租賃系統的出現為的是搭配大眾運輸工具，成為「第一哩」與「最後一哩」的轉乘工具。YouBike於2009年3月試營運，開始於台北市信義區，最初提供11個站點與500輛公共自行車，由於規模較小，未能有效與大眾運輸工具做有效結合運用，導致周轉率不佳。2012年市政府與巨大機械正式簽約，預計在七年內建置163個站點，提供5350輛公共自行車，開啟了YouBike在公共自行車新的歷史。
由於YouBike初期規模較小，運量較低，不論在維護與調度上都能提供高品質的服務。隨著站點的擴張，要達成高品質的車輛維護，與較佳的借還車體驗，如何提供隨時有車借與隨時可還車的情境，場站流量問題成為本論文研究的重心。
本論文以YouBike為例，我們冀望藉由借還車資訊，搭配影響借還車時的天氣因素與空氣污染品質等搭配3種時間序列特徵，利用隨機森林演算法有效的推算出站點在各時段的流量需求。進而提供一線調度人員具有參考價值的站點車輛需求數據，在基本的天氣與空氣污染資訊輔助下，R2-Score可達到0.7左右，而加入時間序列特徵後的R2-Score可達到0.75，最終經過參數的調整，實驗結果顯示R2-Score可再提升5%-6%達到0.8。因此，無論新進人員或有經驗的調度人員都能透過此系統快速精準的掌握各站點在各時段的車輛需求。

In recent years, energy conservation and carbon reduction have become the mainstay of the development of various countries. In order to reduce carbon emissions and solve the problem of traffic congestion, the public bicycle rental system appeared to be used with public transportation, becoming the "first mile" and "last mile" transfer tools. YouBike started its trial operation in March 2009 and started in Xinyi District, Taipei City. It initially provided 11 stations and 500 public bicycles. Due to its small scale, it could not be effectively combined with public transportation, resulting in poor turnover. In 2012, the Taipei City Hall signed a contract with Giant Manufacturing Co. Ltd.. It is expected to build 163 sites and provide 5,350 public bicycles within seven years, opening YouBike's new history in public bicycles.
Due to the small scale of YouBike in the initial stage and the low volume of transportation, it can provide high-quality services in terms of maintenance and scheduling. With the expansion of the station, to achieve high-quality vehicle maintenance, and a better experience of borrowing and returning, how to provide a situation where the bike can be borrowed and returned at any time, the problem of station traffic has become the focus of this paper.
This paper takes YouBike as an example. We hope to use the information of the rental and return, with the weather factors and air pollution quality that affect the rental and return, with 3 kinds of time series features, and use the random forest algorithm to effectively calculate the traffic demand of bike station in time period. Furthermore, it provides reference value of vehicle demand data for front-line dispatchers. With the assistance of basic weather and air pollution information, the R2-Score can reach about 0.7, and the R2-Score after adding time series features can reach 0.75. The parameter adjustment, experimental results show that R2-Score can be increased by 5%-6% to 0.8. Therefore, both new recruits and experienced dispatchers can use this system to quickly and accurately grasp the vehicle needs of each station at each time.

摘要 i
Abstract ii
目次 iii
圖目錄 v
表目錄 vi
第一章 緒論 1
1.1 研究背景與動機 1
1.2 論文貢獻 3
1.3 論文架構 4
第二章 相關研究 5
2.1 公共自行車的演進 5
2.2 相關文獻研究與討論 6
第三章 研究方法 8
3.1 機器學習概述 8
3.2 機器學習的分類與迴歸 9
3.3 機器學習演算法 9
3.3.1 支持向量迴歸 (Support Vector Regression, SVR) 11
3.3.2 極限梯度提升 (eXtreme Gradient Boosting, XGBoost) 12
3.3.3 隨機森林 (Random Forest, RF) 13
3.4 評估指標 17
3.4.1 均方誤差 (Mean Squared Error, MSE) 17
3.4.2 均方根誤差 (Root Mean Squared Error, RMSE) 18
3.4.3 平均絕對誤差 (Mean Absolute Error, MAE) 18
3.4.4 決定係數R2 (R-Square) 18
第四章 實驗結果與評估 20
4.1 流量預測區域選定 20
4.2 資料收集與預處理 21
4.3 場站借還資料分析 23
4.4 實驗設計 27
第五章 結論與未來展望 37
參考文獻 38

[1]黃俊良。〈臺北市公共自行車系統旅次特性分析〉。碩士論文。淡江大學，台灣，2016。
[2]吳冠毅。「分類式支援向量迴歸結合粒子群演算法預測Youbike車位及車輛數」。碩士論文，臺北市立大學資訊科學系，2018。
[3]Wu, Guan-Yi, and Jui-Chung Hung. "Classified support vector regression with particle swarm optimization to forecast youbike parking lots and number of bikes." Proceedings of the 2017 International Conference on Information Technology. 2017.
[4]Chen, Po-Chuan, et al. "Predicting station level demand in a bike-sharing system using recurrent neural networks." IET Intelligent Transport Systems 14.6 (2020): 554-561.
[5]Li, Cheng-Chung. "An improved model to efficiently transport YouBike." Proceedings of the 11th IEEE International Conference on Networking, Sensing and Control. IEEE, 2014.
[6]Wang, Jenhung, Ching-Hui Tsai, and Pei-Chun Lin. "Applying spatial-temporal analysis and retail location theory to public bikes site selection in Taipei." Transportation Research Part A: Policy and Practice 94 (2016): 45-61.
[7]Wong, Jinn-Tsai, and Chun-Yen Cheng. "Exploring activity patterns of the taipei public bikesharing system." Journal of the Eastern Asia Society for Transportation Studies 11 (2015): 1012-1028.
[8]García-Palomares, Juan Carlos, Javier Gutiérrez, and Marta Latorre. "Optimizing the location of stations in bike-sharing programs: A GIS approach." Applied Geography 35.1-2 (2012): 235-246.
[9]張辰尉。「臺北市公共自行車站點需求分析之研究」。碩士論文，國立政治
[10]大學地政學系。
[11]Lin, Jenn-Rong, and Ta-Hui Yang. "Strategic design of public bicycle sharing systems with service level constraints." Transportation research part E: logistics and transportation review 47.2 (2011): 284-294.
[12]Hu, Shou-Ren, and Chao-Tang Liu. "An optimal location model for a bicycle sharing program with truck dispatching consideration." 17th International IEEE Conference on Intelligent Transportation Systems (ITSC). IEEE, 2014.
[13]Chen, Shang-Yu, and Chung-Cheng Lu. "A model of green acceptance and intentions to use bike-sharing: YouBike users in Taiwan." Networks and Spatial Economics 16.4 (2016): 1103-1124.
[14]Ghanem, Ahmed, et al. "Bike share travel time modeling: San Francisco bay area case study." 2017 5th IEEE International Conference on Models and Technologies for Intelligent Transportation Systems (MT-ITS). IEEE, 2017.
[15]Te Pai, Jen, and Shih Ying Pai. "User behaviour analysis of the public bike system in Taipei." International review for spatial planning and sustainable development 3.2 (2015): 39-52.
[16]賴淑芳. "公共自行車接受度與滿意度研究─ 以臺北微笑單車為例." 運輸學刊 24.3 (2012): 379-405.
[17]Vapnik, Vladimir, Steven E. Golowich, and Alex Smola. "Support vector method for function approximation, regression estimation, and signal processing." Advances in neural information processing systems (1997): 281-287.
[18]Cortes, Corinna, and Vladimir Vapnik. "Support-vector networks Machine learning (pp. 237–297), Vol. 20." Boston, MA: Kluwer Academic Publisher (1995).
[19]Chen, Tianqi, and Carlos Guestrin. "Xgboost: A scalable tree boosting system." Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining. 2016.
[20]“XGBoost參數說明.” [Online]. Available: https://xgboost.readthedocs.io/en/latest/parameter.html
[21]L. Breiman, “Random Forests,” Mach. Learn., vol. 45, no. 1, pp. 5–32, 2001
[22]“Scikit Learn Random Forest Regressor參數說明.” [Online]. Available: https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestRegressor.html
[23]Botchkarev, Alexei. "Performance metrics (error measures) in machine learning regression, forecasting and prognostics: Properties and typology." arXiv preprint arXiv:1809.03006 (2018).
[24]“行政院環境保護署 歷年監測資料下載.” [Online]. Available: https://taqm.epa.gov.tw/taqm/tw/YearlyDataDownload.aspx