A Variable Markovian Based Outlier Detection Method for Multi-Dimensional Sequence over Data Stream

摘要: Nowadays sequence data tends to be multi-dimensional over stream, it has a large state space and arrives at unprecedented speed. It is big challenge design outlier detection method meet the accurate high speed requirements. The traditional methods can't handle effectively as they have poor abilities for modeling, detect timely computational complexity. In this paper we propose variable Markovian based VMOD, which consists of two algorithms: mutual information feature selection algorithm (MIFS), sequential analysis (VMSA). uses MIFS reduce redundant features, VMSA accelerate detection. Through VMOD method, can improve rate similarity measures adopt clustering strategy select modeling through reducing consequently, rate. use random sample index structure model construction complexity, quicken experiments show that effectively, time by least 50% compared with methods.