MEOD: Memory-Efficient Outlier Detection on Streaming Data

MEOD: Memory-Efficient Outlier Detection on Streaming Data


In this paper, a memory-efficient outlier detection (MEOD) approach for streaming data is proposed. The approach uses a local correlation integral (LOCI) algorithm for outlier detection, finding the outlier based on the density of neighboring points defined by a given radius. The radius value detection problem is converted into an optimization problem. The radius value is determined using a particle swarm optimization (PSO)-based approach. The results of the MEOD technique application are compared with existing approaches in terms of memory, time, and accuracy, such as the memory-efficient incremental local outlier factor (MiLOF) detection technique. The MEOD technique finds outlier points similar to MiLOF with nearly equal accuracy but requires less memory for processing.


  • Ankita Karale
  • Milena Lazarova
  • Pavlina Koleva
  • Vladimir Poulkov


Symmetry, 2021, 13(3), 458