I-IncLOF: Improved incremental local outlier detection for data streams

Data streams outlier mining is an important and active research issue in anomaly detection. Most of existing methods are more suitable for static data, since algorithms have all data available at time of detection. But, as data streams evolve during the time, traditional methods cannot perform well on them. Therefore, because of dynamic nature of data streams, evaluating objects as outlier when they arrive, although meaningful, often can lead us to a wrong decision. In this paper an Improved Incremental LOF algorithm is proposed. The proposed algorithm considers a sliding window that lets data profiles update during the window and then declares them as outlier/inlier, therefore it can significantly distinct outliers from new data behavior. In addition, I-incLOF declares that there is no need for rerunning deletion algorithm when an outliers is founded, we just do not consider them in the new points neighbors. Our experimental results show that the proposed improved incLOF algorithm was successful in reducing false-positive rate with no additional computational cost.