一种新的聚类组合算法

A new algorithm for clustering combination

提出一种新的基于非负矩阵分解(NMF)方法的聚类组合算法(NMFCCA).该算法首先采用K-均值算法作为基聚类器,然后使用NMF方法从基聚类器输出结果中提取数据对象的关键特征,最后在关键特征空间中划分数据对象,生成最终结果.在人工数据集和真实数据集上的实验表明,所提出的算法是有效可行的.

A new algorithm based on non-negative matrix factorization（NMF） is proposed.It first generates multiple clusterers by K-means algorithm.Then the features of the data objects are extracted by utilizing NMF method.Finally,a single consolidated clustering is built up by dividing the data objects in feature space.According to experiments based on artificial data sets and real data sets,the result shows that the proposed algorithm is feasible and effective.