Abstract: The similarity measurement among data is important for further analysis of the data set. Aiming at the similarity measurement of high dimensional data, the paper put forward a new method based on subspace. After dividing high dimensional space into grids, and computing the similarity among data in proper subspaces, the disturbance from the curse of dimensionality can be abated efficiently under the proper dividing parameters.

Key words: High dimensional data, Curse of dimcnsionality, Grid-based dividing, Subspace, Similarity mcasurcmcnt