Since the clustering effect is poor because the previously-proposed incremental clustering method converts the multi-dimensional large dataset into the one-dimensional large dataset directly,a new incremental clustering method for large da-taset is put forward. In order to obtain the high clustering efficiency,the incremental clustering step of the large dataset was sim-plified while highly maintaining the original data dimensions to construct a large data processing set. The local incremental clus-tering is performed for the logo samples in the set. The large data with failed clustering is distributed into the local incremental clustering results equally,and its fault coordinate is detected with Gaussian probability density function and coordinate evidence theory and modified. The final incremental clustering results are obtained. The experiment results prove that the proposed method has superior clustering effect and clustering efficiency.%以往提出的面向大数据集的递增聚类方法直接将多维度的大数据集转换成一维大数据集,导致聚类成果不佳,故提出面向大数据集的递增聚类新方法.为取得高聚类效率,在高度保留原始数据维度的情况下,简化了大数据集递增聚类步骤,构建出大数据处理集合,对集合中的标志样本进行局部递增聚类,将未能成功聚类的大数据平均分配到局部递增聚类结果中,使用高斯概率密度函数和证据理论检测其中的错误坐标并进行改正,获取最终的递增聚类结果.实验结果证明该方法具有优越的聚类成果和聚类效率.
展开▼
