iSAX – The Required Theory

Now that we know all about SAX, including normalization and computing the SAX representation of a subsequence, it is time to learn the theory behind the iSAX index, which, at the time of writing, is considered one of the best time-series indexes. Improved versions of iSAX that make iSAX faster and more compact exist, but the core ideas remain the same.

As you might have guessed from its name, iSAX depends on SAX in some way. Put simply, the keys to every iSAX index are SAX representations. Therefore, searching in an iSAX index depends on SAX representations.

At this point, I believe it would be good to provide more information about iSAX to help you while reading this chapter. An iSAX index is a tree-like structure where the root, and only the root, can have multiple children, and all the children of the root are binary trees underneath. Additionally, to create an iSAX index, we need a time series and a threshold value, which is the maximum number...