Mining VIP Based on an Improved Hierarchical Clustering Method

Traditional hierarchical clustering (HC) methods which always fail to deal with very large databases or high dimensional spaces. In the paper, arithmetic mean, arithmetic mean ratios and so on are introduced. Variable trend, data preprocessing for database based on variable trend, and mining VIP (variable important in M/Z, or called typical characteristic) based on an improved hierarchical clustering method is put forward. It was proved to be feasible and effective to mining VIP according necessity after tested.