Time series clustering how to#

This papers focuses on the second type of time series clustering, and makes the disruptive claim that clustering of time series subsequences is meaningless!

Subsequence clustering: When the original data is one long time series that needs to be broken into parts to do clustering on those parts.

Whole clustering: When you have several time series which can come from different machines and want to compare them.

There are two types of clustering when dealing with time series data:

Luckily, I found this paper from Eamonn Keogh & Jessica Lin, which is really interesting (and I encorage you to read), but for those of you who don’t have time (or don’t feel like it) I decided to sum it up in a post.

Time series clustering how to#

Recently, I’ve started doing some research in how to apply clustering techniques to time series and I was surprise to find very little posts about the topic (specially comparing it to the number of posts I found when I wanted to learn about LSTMs).

Don’t make this mistake when clustering time series data!ĭisclaimer: This is not original work, the aim of this post is to spread and make more accesible the content of the paper “ Clustering of Time Series Subsequences is Meaningless: Implications for Previous and Future Research” by Eamonn Keogh & Jessica Lin.