投影寻踪动态聚类模型研究及其在洪灾评定中的应用

Study on Projection Pursuit Dynamic Cluster Model and Its Application to Flood Disaster Evaluation

投影寻踪动态聚类模型是在投影寻踪聚类模型基础上发展起来的一种改进模型,该模型在消除原模型中密度窗宽确定的人为任意性的同时,把寻优过程和聚类结果有机地联系起来,具有稳定性好、客观性强以及分类结果明确等特点.洪水灾情的评定结果表明,投影寻踪动态聚类模型不仅能更好地实现投影寻踪的建模思想,取得令人满意的实际应用效果,也为同类问题的解决提供了一种新途径.

Through penetrating analysis of projection pursuit cluster model, an improved model （projection pursuit dynamic cluster model） is proposed. In this model, the subjectivity of the density window radius is effectively eliminated, and at the same time optimization process and cluster result are closely integrated. So projection pursuit dynamic cluster model is robust, and its evaluation is more objective, additionally, its cluster result is also more definite. Finally this model is applied to study flood disaster evaluation, and the tests show that it can not only realize the essence of projection pursuit model and obtain a very satisfying practical result, but also provide a new approach for solution of similar problems.