Time series of different lengths

[atombysx]

Hi,

I'm tempted to use this pipeline after reading the paper.

But my multivariate time series data are at different lengths in time and the pipeline seems requiring the array to be the same size.

Is it possible to extract same features for clustering from MTS at different lengths any way? Or the features are very time-sensitive?

[FrancescoDelBuono]

Thank you for considering Time2Feat.

Currently, for performance reasons, we have decided to enforce a constraint that requires all time series to have the same length, and this issue cannot be resolved by padding with null values.

Theoretically, the model can be adapted to handle time series of variable lengths. If you have at most a limited number of different lengths, I can develop an optimized version in the next few weeks that will accommodate variable lengths

[atombysx]

Thanks for the reply!

For my datasets and analysis, I will do clustering on 100-200 sets of multivariate time seris (5-10 variables). Within each set, the time series lengths are the same among the variables.

Would this be too many sets?

[FrancescoDelBuono]

I believe it is possible. I will try to implement a solution next week.