The data generated from some important applications is inherently uncertain, such as sensor networks and mobile object tracking. Firstly, the traditional agglomerative hierarchical clustering algorithm is extended to uncertain algorithm (U-AHC), which can be handled uncertain data. Secondly, rough table instances of candidate co-location patterns are computed based on the results clustered by U-AHC, and candidate co-location patterns whose rough participation index values are smaller than minimum prevalence threshold are pruned. Then, for the remained candidate co-location patterns, the real table instances are generated and some of them are pruned dynamically. Finally, the experimental results demonstrate that our algorithms are effective and efficient.%不确定数据在一些重要应用领域中是固有存在的,如传感器网络和移动物体追踪等.如何快速、方便、有效地从不确定数据库中发现潜在的、有价值的和人们感兴趣的信息变得越来越重要.首先,把传统的凝聚层次聚类算法(AHC)扩展到不确定的凝聚层次聚类算法(U-AHC),然后在聚类结果的基础上计算候选co-location模式的粗表实例,并对参与度小于最小参与度阈值的候选模式进行剪枝.接着展开其粗表实例并动态地实施剪枝,最后生成频繁的co-location模式.实验证明这个算法是正确的,而且效率较高.
展开▼
