Time series discord discovery using WAT algorithm and iSAX representation

Abstract

Among several existing algorithms proposed to solve the problem of time series discord discovery, HOT SAX and WAT are two widely used algorithms. Especially, WAT can make use of the multi-resolution property in Haar wavelet transform. In this work, we employ state-of-the-art iSAX representation rather than SAX representation in WAT algorithm. To apply iSAX in WAT algorithm, we have to devise two new auxiliary functions and also modify iSAX index structure to adapt Haar transform that is used in WAT algorithm. We empirically evaluate our algorithm with a set of experiments. Experimental results show that WATiSAX algorithm is more effective than original WAT algorithm.