Book Image

Time Series Indexing

By : Mihalis Tsoukalos
Book Image

Time Series Indexing

By: Mihalis Tsoukalos

Overview of this book

Time series are everywhere, ranging from financial data and system metrics to weather stations and medical records. Being able to access, search, and compare time series data quickly is essential, and this comprehensive guide enables you to do just that by helping you explore SAX representation and the most effective time series index, iSAX. The book begins by teaching you about the implementation of SAX representation in Python as well as the iSAX index, along with the required theory sourced from academic research papers. The chapters are filled with figures and plots to help you follow the presented topics and understand key concepts easily. But what makes this book really great is that it contains the right amount of knowledge about time series indexing using the right amount of theory and practice so that you can work with time series and develop time series indexes successfully. Additionally, the presented code can be easily ported to any other modern programming language, such as Swift, Java, C, C++, Ruby, Kotlin, Go, Rust, and JavaScript. By the end of this book, you'll have learned how to harness the power of iSAX and SAX representation to efficiently index and analyze time series data and will be equipped to develop your own time series indexes and effectively work with time series data.
Table of Contents (11 chapters)

Manually constructing an iSAX index

In this section, we are going to manually create a small iSAX index. For a better understanding of the process, we are going to present all the steps and describe all the required computations.

If you recall from earlier on in this chapter, the steps for creating an iSAX index can be described as follows:

  1. Separate a time series into subsequences based on the given sliding window size.
  2. For each subsequence, calculate its SAX representation, based on the given parameters.
  3. Begin inserting the subsequences of the time series into the iSAX index. In the beginning, all iSAX nodes are terminal nodes, apart from the root.
  4. Once a terminal node is full – the threshold value has been reached – split that node by increasing the cardinality of one of its segments and create two new terminal nodes.
  5. The original terminal node becomes an inner node, which is now the father of the new newly created terminal nodes.
  6. Split...