兴趣子空间挖掘算法在高维数据聚类中的应用

Application of Interesting Subspace Mining Algorithm in High-dimensional Data Clustering

给出了兴趣子空间的定义,采用基于Chernoff-Hoeffding边界,带回溯的深度优先搜索算法来挖掘最大兴趣子空间,并运用高维真实数据和合成数据检验算法的有效性。高维数据的挖掘面临着数据分布的稀疏性和特征空间的相交性所带来的挑战。

Based on Chemoff-Hoeffding bound, this paper adopts a novel mining algorithm of depth-first search with backtracking to mine interesting subspace, and testfies the effectiveness by using synthetic and real data. High-dimensional data mining faces the challengers of distributed data sparsity and overlapping feature subspace.