The Spatial Data Mining Based On Geological Hazard Data Warehouse

China is a geological disaster-prone country. Each year geological disasters happened with countless traumas to the national economy and people’s living. In recent years, the earthquakes, avalanches, landslides, mudslides and other geological disasters with the risk assessment has become one of the main issues of general interest. According to the2012China Land and Resources Bulletin:In2012,there are14322geological disasters happened, of which10888landslide,2088collapse,922mudslides,347ground subsidence,55ground fissures,22cases of land subsidence.292people were killed,83people were missing,259people were injured in the catastrophe. Compared with the previous year, the number of occurrences has a decrease of8.6%, while casualties and missing people has an increase of35.4%.Geological disasters caused direct economic losses of52.8billion yuan, an increase of31.7%over the previous year. Geological disasters caused the indirect loss is immeasurable, brought a huge loss to the country and the people’s lives and property, and even affected the country’s technical and economic policies. The prevention and mitigation of landslides and other geological disasters is significant, and directly related to the economic development and social stability. Each year the economic loss caused only by the landslide is up to billions of dollars.Therefore, the study of landslide risk has become a hot issue today.However, subject to economic conditions as well as the limitations of the understanding and management level, we can only use the limited resources to take measures to deal with the frequent occurrence of the landslide hazard. The response of management departments and research institutes in the geological disaster is often lagging. The measurement of prevention and reduction is rarely effective, and also in a narrow range. At the same time, while the masses of people have been generally aware of the risks posed by the geological disasters, but in the practice of the process of disaster prevention and mitigation, the broad participation is still very limited. The main reason for this situation is the geological disasters location, time, size, and its great uncertainty. There are still many deficiencies. The lack of quantitative basis in the development of disaster prevention and mitigation programs can not lead to a definitive decision-making. It all depends on the improvement of geological hazard assessment and risk evaluation level. However, both in theory and in practice, China has not yet formed a practical regional geological hazard assessment and management system.Since2003, geological disaster monitoring and early warning has been implemented in various districts counties in Three Gorges Reservoir. The monthly monitoring has accumulated a massive data, including the water level, rainfall, and landslide displacement data. How to effectively process the mass monitoring data has became the current problem.Based on the above problem, the author obtained the following results though the research:(1)By calculating the Pearson coefficient between the bivariate, the initial quantitative understanding is formed with the landslide deformation of displacement on various factors. Then, by the Apriori association rules analysis, it identifies the relationship of landslide deformation displacement with the dam water level, water level fluctuation speed, as well as natural rainfall.(2) The fitting curve of multiple regression model of Bazimen and Shuping landslide displacement is close to the actual displacement trends. Neural network fitted displacement good in the flood season, it’s just multiple regression’s complementary.(3) In this paper, landslide displacement prediction is solved by artificial neural network. The prediction accuracy of Bazimen and Shuping landslide displacement is less than ideal, to achieve in the short-term forecasts, the accuracy of the model has yet to be further improved.(4) Based on the method of time series, the landslide displacement is decomposed into trend term and periodic term. And set appropriate ARIMA model for water level, rainfall and landslide displacement.The innovation of this paper is to put forward a series of landslide data mining methods, maximized the utilization of historical data. On an experimental basis, it has got some new understanding of the landslide.