本研究利用地理資訊系統技術，結合空間決策支援系統結合決策支援理論，以非監督式最佳化的群聚觀念，代替監督式的分類學習，解決坡地災害的決策效能。過去進行災害分析時，多採取資
料維度縮減的分類技術（dimensional reduction），它可以協助使用者從大量的資料中分析資料間的結構、簡化資料的複雜性。本研究採其優點並以聚類分析(K-means)與粒子群最佳化（Particle Swarm Optimization）技術（簡稱為 KPSO），在知識界限圖內選取最具代表
性的臨界點（切割點）及建立知識規則，並以離散化算法，求出重要核心因子，獲得門檻值（threshold） 。研究案例為雪霸地區崩塌地，在取得該區資料庫中的之地文、水文因子、地形因子、植生因子等相關因子後，運用 KPSO 分析探討崩塌之影響機制及崩塌地潛感分布。同時本研究也同步以 SOM( Self Organization Map, 自組特徵映射網路) 進行分類的研判比較，由結果中發現，所提供之最佳化技術，有助於掌握崩塌發生之詮釋空間資訊，進而建置完整之決策支援系統。

The SDSS (Spatial Decision Support System) integrates GIS and Spatial Decision System to efficiently resolve the decision on landslide occurrences. Supervised learning system with optimal clustering concept is applied on the study area –National Shei Pa Park. Dimensional reduction in Data Mining techniques is widely used in this field of this study. Hence, the present study used a hybrid model of K-means and Particle Swarm Optimization to resolve the complexity on the various attributes of data regarding to the landslide factors. That is, an unsupervised clustering approach is used to substitute the conventional supervised approach. As part of this study, a series of extensive in-situ data of soil condition is investigated. Those data contain geomorphological, hydraulic conditions and vegetation distribution from Spatial-Information (GIS and RS) database. Also, the well-known technique of SOM (Self organization Map) is used for comparison. Based on the well-developed classifier, our SDSS can spatially depict the landslide zone successfully.