Across 124 time-series classification datasets, six open-source feature sets perform mostly equivalently, with tsfresh winning most often and simple quantile/FFT baselines competitive on several problems.
The Journal of Machine Learning Research7, 1–30 (2006)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ME 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Statistical comparisons of time-series feature sets on classification tasks
Across 124 time-series classification datasets, six open-source feature sets perform mostly equivalently, with tsfresh winning most often and simple quantile/FFT baselines competitive on several problems.