ImputePilot
Time Series Imputation
Upload your dataset and train a recommendation model
Upload time series data. The paper uses 67K series from 107 datasets.
Click to upload training data
.zip, .csv, .txt, .tsv
Shape-based clustering using cross-correlation with merge/move refinement.
Benchmark algorithms on cluster representatives, propagate labels within clusters.
Extract 808 features using 3 extractors (Catch22 + TSFresh + Topological)
Lightweight canonical time series features
Comprehensive statistical features library
Persistence diagrams from TDA
Pipeline competition with two-phase pruning.
Upload your dataset and train a recommendation model
Click to upload your data
.zip, .csv, .txt, .tsv
Extract 808 features using 3 extractors (Catch22 + TSFresh + Topological)
Lightweight canonical time series features
Comprehensive statistical features library
Persistence diagrams from TDA
Each winning pipeline votes with probability; average determines winner
| Algorithm | Pipeline1 | Pipeline2 | Pipeline3 | Average |
|---|---|---|---|---|
Select baselines to compare. Train them first if not yet trained.
Evaluate how imputation quality will impact downstream tasks
Predict future values
Categorize time series