基于单元的快速的大数据集离群数据挖掘算法

提出基于单元的快速的大数据集离群数据挖掘算法，用聚簇技术对数据进行预处理，然后将数据放入合适的空间单元并对非空单元使用维单元树（cell dimension tree，CD-tree）进行索引，数据集中大部分位于高密度区且与离群数据无关的数据将会被过滤掉，从而避免了大量不必要的计算。实验表明，该算法能快速准确地从大数据集中挖掘出离群数据，并提高离群数据的检测速度。

Fast outlier data mining algorithm based on cell in large datasets

The paper proposed a fast cell-based algorithm for outlier detection in large datasets(short for FOMABCLD). The algorithm applied cluster technique to preprocesse data, and placed data into the appropriate cells based on their values and indexed the non-empty cells with Cell Dimension Tree. A majority part of data located in high density cells and had no nearness relationship with outliers is filtered, which avoided large useless computations. The experiment show that FOMABCLD can mine outlier data from large datasets fast and accurately, and the speed of detecting outliers is increased.