Discussion about Threshold of Set-difference in Some Clustering Algorithm

Using CABOFSV to cluster, whether b, the beginning parameter, threshold value of set-square-difference, also named up-bound of a cluster, is reasonable or not is fatal to clustering results. In this paper, how to determine the threshold value of set-square-difference in CABOSFV algorithm is deeply studied. Then, the method of how to determine threshold value of set-square-difference is put forward and is expressed by a formula, b = (Sigma_i1 = 1 ^{n - k} Sigma_i2=i₁ + 1 ^{n - k + 1}...Sigma _ik = i_{k - 1} + 1^{n - 1} d(i₁, i ₂,...i_k))/C_n ^k, and a group of data is calculated by this method. The calculating results indicate that this threshold is reasonable for CABOSFV because clustering results are controlled by people and ensured correctly and effectively for this method can fix the least number of objects in one cluster