UCR Time Series Classification Archive (128 datasets)

Canonical time-series classification benchmark: a suite of 128 univariate time-series datasets (112-dataset subset commonly used for comparison). Methods are ranked by mean accuracy (and mean rank) across all datasets.