Systematic Feature Engineering for Time-Series Data Mining (SFE-TSDM)

Background

Time series are one of the most common data types in science, engineering, medicine, and economics. In many applications, not the sequential time-series values themselves, but their properties (e.g., autocorrelation structure, entropy, outliers, etc.) are important for analyzing and understanding the respective systems from which they have been recorded. These time-series features have the benefit that they are interpretable, provide valuable insights for domain experts and support explainable machine-learning models.

In recent years, a variety of time-series feature extraction software packages have been developed, including hctsa (Matlab), tsfresh (Python), and feasts (R).

These packages allow users to compute large numbers of univariate time-series features (e.g., up to 7700 features per time series in hctsa) by providing implementations of a wide variety of time-series analysis algorithms, including those developed in statistics, signal processing, time-series analysis, and non-linear dynamics.

Submissions to the workshop on Systematic Feature Engineering for Time-Series Data Mining

The workshop on Systematic Feature Engineering for Time-Series Data Mining is organized as part of the 21st IEEE International Conference on Data Mining, which will be held from 7-10 December 2021 in Auckland, New Zealand. Due to the changing circumstances derived from the COVID-19 pandemic, the format may be updated to virtual only or a combination of virtual and in-person participation. Updates will be posted on this website.

The workshop seeks to connect the data-mining community with researchers and industry professionals using time-series feature engineering.

The workshop organizers are seeking contributions on systematic time-series feature engineering (STSFE), including:

• time-series data mining,
• novel algorithms for time-series feature extraction, including algorithms using neural networks,
• explainable machine learning on STSFE,
• time-series feature-based dimensionality reduction, classification, and regression, and
• evaluating the performance of time-series feature sets,
• constructing reduced feature sets,
• pattern recognition on time series,
• domain-specific and industry applications,
• event sequences and other types of ordered data like language time-series.

Authors are invited to submit original papers, which have not been published elsewhere and which are not currently under consideration for another journal, conference or workshop.

Paper submissions should be limited to a maximum of ten (10) pages, in the IEEE 2-column format (https://icdm2021.auckland.ac.nz/cfp/), including the bibliography and any possible appendices. Submissions longer than 10 pages will be rejected without review. All submissions will be triple-blind reviewed by the Workshop Committee on the basis of technical quality, relevance to scope of the workshop, originality, significance, and clarity. Manuscripts must be submitted electronically in the online submission system https://wi-lab.com/cyberchair/2021/icdm21/scripts/submit.php?subarea=S21. We do not accept email submissions.

The accepted workshop papers will be published in the dedicated ICDMW proceedings published by the IEEE Computer Society Press.

If you have any questions, feel free to contact the workshop organizers:

• Ben Fulcher
• Andreas W. Kempa-Liehr

Key Dates

• Workshop paper submissions: September 03, 2021 September 10, 2021.
• Workshop paper notifications: September 24, 2021.
• Camera-ready deadline and copyright forms: October 1, 2021.
• Workshop: December 7, 2021.

All dates are provided in 11:59pm Pacific Daylight Time (PDT).