- 期刊
應用羅吉斯迴歸預測臺中市山坡地違規
Prediction of Violations of Slopeland Regulations in Taichung City Through Logistic Regression Analysis
摘要
本研究以臺中市山坡地為研究對象，首先蒐集研究區之基本資料與山坡地的違規資訊，先採用核密度分析法，針對臺中縣、市合併前後，分析山坡地違規的熱區之空間分布，選取違規分析之代表因子，包括：道路距、社區活動中心距、觀光景點距及坡度，採用羅吉斯迴歸法預測山坡地違規分佈，並針對不同網格解析度對結果之影響進行評析。研究結果顯示，臺中山坡地違規熱區在縣市合併前為南屯區、潭子區及北屯等三個區域，合併後另增加沙鹿區、太平區、霧峰區及東勢區，前述所示山坡地違規圍繞在都會區與重要交通樞紐區的周圍，違規聚集的趨勢隨著不同時期而改變；採用羅吉斯迴歸模式進行違規預測，其AUC值均達到80%以上，模式具有優良的鑑別力。最後本研究建立違規潛勢分佈圖，未來可針對違規高潛勢區域，可透過增加巡邏、加強法治教育及設立山坡地違規使用檢舉告示牌等預防措施，降低違規行為與減少可能衍生之災害。
並列摘要
This study investigated hillslopes in Taichung City, collected basic data pertaining to the study areas, and examined the violations of hillslope regulations in these areas in recent years. Kernel density estimation was first performed to determine the spatial distribution of violation hotspots in Taichung City before and after the merging of Taichung City and Taichung County to form the Taichung City special municipality in 2010. The violation analysis included the distance to a road, the distance to a community activity center, and the distance to a tourist attraction and a slope. Subsequently, logistic regression was performed to establish a model for predicting hotspots for hillslope-related violations. The effects of grid resolution analysis on the model's results were investigated. The results indicated that before the city-county merger, the hotspots for hillslope-related violations were identified in three districts of Taichung City, namely Nantun, Tanzi, and Beitun districts. After the city–county merger, hotspots were also identified in the districts of Shalu, Taiping, Wufeng, and Dongshi. The violations of hillslope regulations were distributed across the city and its key transportation areas, with the trend for violations changing over time. For model evaluation, the logistic regression model produced accurate predictions, achieving an area under the curve of >80% with the application of multiple grid resolutions. Maps displaying predicted violations were subsequently obtained. The results indicate that introducing preventive measures (e.g., increasing the number of patrols, enhancing education on the regulations, and establishing a violation bulletin) that target the predicted violation areas can prevent future violations and reduce the risk of related disasters.