Detection Algorithm of Local Outlier Based on Variable Dense Cell

Aiming at the problems of low detection rate and poor time performance in the detection of abnormal points in the LOF algorithm, an improved algorithm for local outlier detection of variable dense cell is proposed. The algorithm has been improved in two aspects. One is the dynamic meshing of variable density cells, which can merge adjacent grids with similar data into a large grid, reducing the number of grids. The other is to quickly set some key parameters of dynamic mesh generation. The experimental results show that compared with the outlier detection algorithms in the literature, the improved algorithm has a certain degree of improvement in meshing effect, time performance and detection rate.