請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/68556
標題: | 應用大數據探勘技術探討台灣城市之永續性 Exploring City Sustainability in Taiwan through Data Mining Techniques |
作者: | Ying-Ya Liao 廖盈雅 |
指導教授: | 張斐章 |
關鍵字: | 資料探勘技術,大數據,自組特徵映射組織網路,調適性網路模糊推論系統,永續發展,水-糧食-能源鏈結,溫室氣體排放, Data mining,Big Data,Self-organizing map (SOM),Gamma Test (GT),Adapted Network-Based Fuzzy Inference System (ANFIS),Water-Food-Energy Nexus,Sustainable Development,Greenhouse Gas (GHG), |
出版年 : | 2017 |
學位: | 碩士 |
摘要: | 台灣在過去半個世紀以來,因經濟快速發展以及都市化影響,對於水、糧食及能源的需求日益增加,資源的過度消耗更導致城市負載力(Urban Carrying Capacity)增加及全球暖化等問題,威脅著城市環境與經濟之永續發展。2014年由聯合國於所訂定17項永續發展目標(Sustainable Development Goals, SDGs)之中指出溫室氣體的大幅增加將導致國家社會暴露於嚴重風險下並威脅國家永續發展,加上許多過去研究指出涉及國家永續發展之一為資源的使用情況,因此為了邁向國家城市之永續發展,透過先瞭解城市之水、糧食與能源等資源使用情況及社會經濟條件,並從大數據中找出和溫室氣體排放量之間影響之關鍵因素。
本研究之資料蒐集為環繞淡水河流域之四縣市(分別是新北市、台北市、桃園市、基隆市)和環繞濁水溪流域之兩縣市(彰化縣和雲林縣)等共六縣市之水、糧食、能源、經濟人口、溫室氣體等相關之歷史數據。由考慮聯合國世界環境發展委員會(WCED)提出永續發展中之社會-經濟-環境三環面向,取得其中三環所涉及資料以作此研究樣本以探討資源使用情況與溫室氣體排放量,予以探勘各縣市近二十年來城市之發展趨勢。 研究方法首先以過去二十年歷史資料分析各城市資源使用情況與經濟發展樣貌。接著應用自組特徵映射組織網絡(Self-Organizing Map, SOM) 作為資料探勘工具,依各異質因子之特性作群聚分析,挖掘尋找資料關聯性;接以Gamma Test從聚類結果中篩選出與溫室氣體之間關鍵因子,再以調適性網路模糊推論系統(Adapted Network-Based Fuzzy Inference System, ANFIS)建置溫室氣體排放之推估模式。 根據研究結果顯示:各縣市之城市資源使用情況與經濟發展確實因地方發展型態不同而呈現面貌不同,尤其分別兩大流域發展程度之間有較大差異。由資料探勘技術SOM可有效率地呈現異質資料的特性,SOM之聚類結果顯示,資料特性的表現成三類型,分別為第一型與民生相關、第二型與工業相關以及第三型與農業相關。其中第一型包含:人口密度、人均年所得、人均飲食消費、人均民生用水量以及人均民生用電量,第二型包含:單位面積工業用水量及單位面積工業用電量,第三型包含:單位面積灌溉用水量、單位面積灌溉用電量、單位面積稻米生產量、單位面積蔬菜生產量及單位面積水果生產量。此外,透過Gamma Test檢定所篩選出與溫室氣體相關之關鍵因子分別為人均民生用水量、單位面積工業用電量及單位面積灌溉用水量,並作為ANFIS輸入值。最後ANFIS溫室氣體排放量推估模式結果顯示,由規則3個時的表現最佳,其推估模式之相關係數(CC)於訓練、驗證及測試皆高達0.99,表示模式精準度相當高,具有效率之推估能力。由本研究之大數據探勘分析及建立GHG排放量之推估模式,可從大數據中找出關鍵因子對GHG與城市永續發展間之影響,期望未來可提供水、糧食與能源資源以及城市永續發展管理之參考依據。 Over the last 50 years, the rapid pace of urbanization in Taiwan has resulted in significant social and environmental changes. The increasing demand for water, food and energy, and excessive consumption of resources has led to problems such as urban carrying capacity and global warming. These problems have threatened the sustainable development of a city and its economy. Studies further pointed out that the significant increases of greenhouse gas emissions (GHG) could lead to serious risks for the human society, threatening national sustainable development among the sustainable development goals (SDGs) set by the United Nations in 2014. Many studies also pointed out that effective use of resources is one of the keys for sustainable development of a city. To achieve the sustainability, understanding the use of resources such as water, food and energy, and socio-economy conditions is important. Consequently, this study aims to explore the relationships among resources and the GHG emissions through the data mining and clustering techniques. The objective of the study is to examine key factors of GHG emissions and using the artificial intelligence (AI) soft-computing to discover rule-based relationships among the key factors and the GHG emissions. Information collected in this study included water usage, food production, energy usage, and economic-related data, as well as population and GHG emissions. This study considered 6 cities including New Taipei City, Taipei City, Taoyuan City and Keelung City surrounding the Tamsui River Watershed, as well as Changhua County and Yunlin County surrounding the Jhuoshuei River Watershed. Considering the socioeconomic-environmental sector of sustainable development proposed by the Union Nations World Environment Development Council (WCED), I used the aforementioned long-term heterogeneous data to explore the relationship between resource consumption and GHG emissions. I then assessed the development trend of cities and analyzed the resource consumption and economic development over the past two decades. Following that, I applied the Self-Organizing Map (SOM) to cluster heterogeneous data for exploring the nexus among the resources of water, food, and energy. After the SOM, a Gamma Test (GT) was applied to screen the key factors most related to GHG emissions from the clustering results. Lastly, I put forward an estimation model for GHG emissions by the Adapted Network-Based Fuzzy Inference System (ANFIS). The research results found that economic and resource development trends varied significantly by cities. The results of SOM clustering displayed the trends in three different types representing the domestic, industrial, and agricultural sectors. The domestic sector included population density, per capita annual income, per capita consumption of food, per capita consumption of water and per capita consumption of electricity. The industrial sector included industrial water consumption per unit area and unit area of industrial electricity consumption. The agricultural sector included irrigation water usage per unit area, irrigation electricity consumption per unit area, rice production per unit area, vegetable production per unit area and fruit production per unit area. In addition, the GT results selected three key factors of per capita consumption of water, unit area of industrial electricity consumption, and irrigation water usage per unit area representing major influences from each sector to the GHG emissions. Based on the ANFIS model, results showed excellent prediction accuracy by using the three key factors, with a correlation coefficient (CC) in the training, validation and testing phases up to 0.99 from the rule 3 of membership function. Overall this study demonstrated that through the AI data mining and clustering techniques, the nexus among different resources can be explored. The identified key factors by GT to develop the GHG estimation model using ANFIS can be used to provide useful information for effective and sustainable management on Water-Food-Energy (WFE). |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/68556 |
DOI: | 10.6342/NTU201704031 |
全文授權: | 有償授權 |
顯示於系所單位: | 生物環境系統工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-106-1.pdf 目前未授權公開取用 | 4.58 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。