基于可调整邻域阈值的DBSCAN算法在应急预案分类管理中的应用

Application of DBSCAN algorithm based on adjustable threshold in the emergency plan classification management

针对庞大的预案文本资源分类难的问题,将可调整的邻域阈值Eps取代原有的全局Eps,得到了改进的DBSCAN密度聚类算法.以预案文本间的相似度作为聚类基本的度量属性,将改进的DB-SCAN算法应用于应急预案分类管理,以去除边界.仿真结果证明该方法不仅不影响预案本来的基础分类方式,而且更易于实现,在一定程度上能够缓解噪音点误识别问题,对提高预案文本的重用性和分类的准确率有一定的参考意义.

Aiming at large plan texts resource classification problems, adjustable threshold Eps replaced the original global threshold Eps. An improved DBSCAN clustering algorithm based on density was put for- ward. The similarity between plan texts was taken as measurement attribute. Improved DBSCAN was ap- plied in the field of plan classification to remove the boundary identification error. The simulation results showed that this method not only does not affect the result in basis classification way, but also have certain reference significance to improve accuracy and reusability of classification.