nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Pattern recognition in time series

I have found how clustering techniques can be used in JMP to identify similar patterns:

https://community.jmp.com/t5/Discussions/How-to-determine-clusters-of-semiconductor-wafer-defect-sort/td-p/15051

When analyzing time series data I am often interested in clustering periods where a pattern of interest appears. In the literature there are several techniques and I wonder how JMP/JMP Pro can be used to have some initial results.

Accepted answers

Jeff_Perkinson

I think the Functional Data Explorer platform may help you here. You can save functional principal components for each time series and then cluster based on those principal components.

Here are some tutorials on the Functional Data Explorer:

The What, Where, Why and How of Functional Data Explorer ( 2019-EU-45MP-050 )

The Wide Technical Facility of Functional Data Explorer ( US 2018 226 )

Tutorial - Using Functional Data Explorer to Make Sense of Sensor Data (US 2018 410)

All comments

P_Bartell

Can you share the specific instances/techniques in the literature that you are interested in using? Then compared capabilties in JMP/JMP Pro wrt to these specific instances?

Jeff_Perkinson

I think the Functional Data Explorer platform may help you here. You can save functional principal components for each time series and then cluster based on those principal components.

Here are some tutorials on the Functional Data Explorer:

The What, Where, Why and How of Functional Data Explorer ( 2019-EU-45MP-050 )

The Wide Technical Facility of Functional Data Explorer ( US 2018 226 )

Tutorial - Using Functional Data Explorer to Make Sense of Sensor Data (US 2018 410)

Interesting approach - one can take advantage of the data preprocessing and dynamic time warping within the FDE platform.

This answer from quora is a good summary of different techniques that can be used previous clustering.

src: quora.com/What-is-the-best-way-algorithm-to-detect-a-pattern-in-a-time-series

Cross correlation
Windowed dynamic time warping
Fingerpinting the time series using a combination of statistics
Hidden Markov Models
Recurrent Neural Networks

P_Bartell

In addition to the Functional Data Explorer platform (only available in JMP Pro) you may also want to check out the Time Series platform in JMP (which is also in JMP Pro since JMP capabilities are a subset of JMP Pro capabilties). There are many pattern recognition subplatforms/visualization options nested within the platform. As one of the quora respondents so wisely stated, '...it depends on the pattern your're looking for.' Here's a link to the JMP online documentation for the Time Series platform:

https://www.jmp.com/support/help/14-2/time-series-analysis.shtml#