基於極限梯度提升方法以評估及預測綠色城市

快速的都市化發展，環境汙染、生態失衡問題日益嚴重，導致氣候變遷，極端氣候所帶來的高溫、乾旱、豪雨及野火等威脅，大家開始注重「永續」、「環保」和「低碳」等議題，而如何既保護環境又維持居民的生活品質以成為綠色城市是許多城市都在追求的，因此預測是否為綠色城市及探討綠色指標重要性是很重要的。本研究將參考歐盟Green City Index、European Green Capital Award和聯合國Sustainable Development Goals等指標，利用XGBoost集成式學習模型預測各城市是否達到綠色城市標準已成為綠色城市，選用XGBoost是因為它有產生多個模型並基於前先前模型進行優化的特性，所以模型效果會更好，最後將結果和傳統邏輯回歸所建立之模型進行各指標間的相關性和模型準確率的比較，而結果顯示比較重要的指標有人均溫室氣體排放量跟人均交通所產生的二氧化碳，因此政府或相關部門可以優先制定此方向相關規定以減少排放量，加快達到綠色城市的速度，也可以先進行綠色城市預測後再比對綠色城市數據，針對不同需求進行調整，以改善環境達到綠色城市。

關鍵字

綠色城市；永續發展；綠色城市指標；模型預測；極限梯度提升

並列摘要

With the rapid development of urbanization, environmental pollution and ecological imbalance are becoming more and more serious and leading to climate change. Due to the threats of high temperature, drought, heavy rain and wildfire brought by extreme climate, people are beginning to pay attention to "sustainability", "environmental protection" and "low carbon" and other issues. How to protect the environment and maintain the quality of life of residents becomes what cities focus on. Therefore, it is very important to predict green cities and evaluate the green indicators. This study considers indicators such as the EU Green City Index, the EU European Green Capital Award, and the United Nations Sustainable Development Goals. Extreme Gradient Boosting is used to integrate learning model to predict whether each city is a green city and to generate multiple models, which are optimized based on previous models. The characteristics of the model have impact on the model’s performally. Finally, the results were compared with the model established by the traditional logistic regression for the correlation and accuracy of each indicator. The results show that the more important indicators are “Greenhouse Gas Emissions per Person” and “per capita Transport Emissopns from Transport”. The government or relevant departments can give priority to formulate relevant regulations based on the two indicators to reduce emissions. Alternatively, make green city predictions first and then compare green city data, and adjust according to different needs to improve the environment and achieve green cities.