Construction of knowledge-based spatial decision support system for landslide mapping using fuzzy clustering and KPSO analysis

摘要

A spatial decision support system, incorporating a geographic information system and a data interpreter based on Data Mining, is developed to analyze the landslide distributions and locations. Satellite remote sensing (RS) can offer an advancing scientific-based knowledge of the landslide problem that, directly and instantly adopt to present the disaster area. However, integrating the RS data into a decision system seems quite difficult. Therefore, this study is decided to develop unsupervised clustering techniques with an improved translation platform to extract some image knowledge on landslide occurrence. More specifically, this study used spatial information technology to attain the vegetation cover and landforms. Conditioning factors are also adopted to attain the further vegetation information. Then, two parallel spatial decision support systems are generated: (a) fuzzy c-means (FCM) is used to analyze the feature of their attributes; (b) the KPSO (k-means + Particle Swam Optimization) is used to approach a parallel study of FCM. Various levels of m values (different fuzzy degrees) are presented with regards to the accuracy of the image classification. While m = 2 the accuracy is 81 % which is lower than m = 3.6 the accuracy is 86 %. Similar results are obtained through KPSO and verifications are made. Finally, the EKTP (Expert Knowledge Translation Platform) is applied to enhance the performance of accuracy.