透過您的圖書館登入
IP:3.133.144.217
  • 期刊

應用巨量資料空間資訊系統之建置:利用粒子群最佳化分析建構崩塌地

The Study of Big Data Analysis on Spatial Information System: Particle Swarm Optimization Approach for Landslide Mapping

摘要


空間決策支援系統結合地理資訊系統技術與決策支援理論,可提升對空間問題的決策效能,過去分析崩塌地的研究著重於統計和模擬,本研究嘗試以非監督式最佳化的群聚理論,進行崩塌地的空間分析。巨量數據因為其維度很高,在收集資料時必須將沒有用的屬性進行刪減,因此資料維度縮減技術(dimensional reduction)進行,可以協助使用者從大量的資料中挖掘資料間的結構、簡化資料的複雜性,以解決空間分析的困難性。本研究嘗試以聚類分析(K-means)結合粒子群最佳化(Particle Swarm Optimization)在知識界限圖內,以雪霸崩塌地的資料庫中,選取最具代表性的臨界點(切割點),利用粒子群最佳化建立知識規則,依其將真值量化後之離散化算法,求出重要核心因子,獲得門檻值(threshold)。即以研究區內崩塌地的空間分布進行運用KPSO(K-means + Particle Swarm Optimization)分析,並取得崩塌地之地文、水文因子、地形因子、植生因子等相關因子探討,崩塌之影響機制及崩塌地潛感分布,期能有助於掌握崩塌發生之詮釋空間資訊,進而建置完整之崩塌地決策支援系統。

並列摘要


The spatial decision support system combines geographic information system technology and decision support theory to improve the decision-making efficiency of spatial problems. The past researches on the analysis of landslide are focused on statistics and simulation. This research attempts to use unsupervised optimization clustering theory to carry out the spatial analysis of landslide. Because of its high dimensionality, large amounts of data must be deleted when collecting data. Therefore, dimensional reduction technology can dig out the structure of data and simplify the data from a large amount of data to solve the difficulty of spatial analysis. However, this research attempts to use cluster analysis (K-means) combined with particle swarm optimization (Particle Swarm Optimization) to select the most representative critical point (cutting Point). The particle swarm optimization to establish knowledge rules, according to the discretization algorithm after the truth value is quantified as well as find the important core factors and then finally obtain the threshold. That is, the hybrid model of KPSO (K-means + Particle Swarm Optimization) analysis based on the spatial distribution of the landslide in the study area. The geological, hydrological factors, topographical factors, vegetation factors and other related factors of the landslide is then applied to discuss the impact mechanism. The potential distribution of the landslide can help to grasp the interpretive spatial information and then build a complete landslide decision support system.

參考文獻


雷祖強(2004):遙測與地景生態分析應用於雪霸國家公園之研究。內政部營建署雪霸國家公園管理處委託研究。
萬絢、王吉成、雷祖強、周天穎(2007):以空間資訊技術與布林粗糙集合分析雪霸崩塌地石礫土與壤土之差異性研究。台灣地理資訊學刊,第五期,1-15。
陳慶逸、余繁(2004):應用於資料探勘之演化式群聚分析技術。智慧型知識經濟暨第二屆演化代計算應用研討會,pp.153-161。台北。
陳怡睿、謝舜傑、陳信達(2005):應用知識庫分類法判釋 SPOT衛星影像坡地崩塌之研究。台灣地理資訊學會年會暨學術研討會。台中。
曾憲雄、蔡秀滿、蘇東興、曾秋蓉、王慶堯(2006):資料探勘Data Mining。臺北市:旗標出版社。pp.4-2~4-25,2006。

延伸閱讀