研究期間：10108~10207;Taiwan is located on the west Pacific where typhoons and other extreme weathers occur frequently. The fractural geological conditions make the land very unstable. Therefore, landslides and debris flows are constantly triggered by heavy rainfall and earthquakes in Taiwan, especially in mountainous and hilly regions. Consequently, understanding the relationship between landslide and spatial factors has become an important issue in hazard mitigation. Remote sensing and spatial analysis are effective methods for monitoring landslide hazards. However, to achieve landslide forecast and risk assessment will require in-depth analysis of landslides and various causative factors. The progress of spatial technologies has led to the increase of data volume and complex data characteristics. It has become a challenging issue for effectively extract useful information and knowledge from the vast amount of heterogeneous spatial data sets. In this regards, data mining seems to be an effective approach. However, because of the uncertainty in spatial data and analysis, directly applying existing data mining techniques for landslide analysis may not produce accurate results. It is necessary to develop novel spatial data mining algorithms based on the characteristics of spatial data and spatial analysis in order to achieve effective landslide information extraction and knowledge discovery from complicated data sets and further establish reliable landslide risk assessment mechanisms for better hazard prevention and mitigation. This project proposes to undertake the advanced research and development of integrated data mining algorithms that are based on spatial uncertainty and specifically designed for landslide hazard risk assessment. Important topics of the research include: understanding of basic spatial uncertainty and spatial data mining, optimization and integration of data mining algorithms for landslide knowledge discovery, landslide hazard risk assessment. In addition, this research will also utilize the developed integrated data mining algorithms to extract landslide causative factors in the Shimen reservoir watershed to establish a landslide prediction mechanism and perform hazard risk assessment of the study site.
