Clustering high dimensional data streams based on N-most interesting itemsets mining

The key for clustering high dimensional data streams is finding dense units. Traditional methods apply frequent itemsets mining for finding dense units. Since these methods are not able to differentiate the density of units in subspaces with different dimensions, it is not in favor of finding dense units in the sparse subspace or the higher-dimension subspace. In this paper, we propose an algorithm, called CBNI (Clustering high dimensional data streams Based on N-most interesting Itemsets), which finds dense units based on N-most interesting itemsets mining and can solve this problem. The experimental results show that the CBNI algorithm performs better in terms of the scalability with dimensionality, the scalability with the number of points in dataset, and the cluster purity.