Implementing the MPdist calculation in Python
In this section, we will discuss two ways to approximately compute MPdist with the help of iSAX.
The first way is much simpler than the second one and is slightly based on the approximate calculation of the Matrix Profile. We take each subsequence from the first time series, and we match it with a terminal node with the same SAX representation from the iSAX index of the second time series in order to get the approximate nearest neighbor – if a subsequence does not have a match based on its SAX representation, we ignore that subsequence. So, in this case, we do not join iSAX indexes, which makes the process much slower – our experiments are going to show how much slower this technique is.
For the second way, we just use the similarity join of two iSAX indexes, which we first saw in Chapter 5.
The next subsection shows the implementation of the first technique.
Using the approximate Matrix Profile way
Although...