Outlier detection is one of the major problems in modern applications. Specially, detecting outliers for streaming applications, as data can dynamically change in subtle ways following changes in the underlying infrastructure. Due to the evolution in data in ratio of data generated every second and velocity, detecting outliers in these types of data becomes a very challenging task. This makes processing the whole data one time is impossible. In this paper we propose a parallel window based local outlier detection (PWLOD) algorithm that can detect outliers in real time using the sliding window algorithm and partition each window among several processing nodes. Each processing node process its portion of window using Local Outlier Factor algorithm and send the results to the master node which collects the results and process them to select the outliers. The experimental results show that the proposed algorithm has better execution time and accuracy than the state-of-the-art algorithms. Information Sciences Letters An International Journal
Sakr, Mohamed; Atwa, Walid; and Keshk, Arabi
"Parallel outlier detection in real time data streams,"
Information Sciences Letters: Vol. 9
, Article 8.
Available at: https://digitalcommons.aaru.edu.jo/isl/vol9/iss3/8