Datasets:

the-piyushgoel
/

drift-forecast-results

Tags:

License:

Dataset card Data Studio Files Files and versions

xet

Community

Dataset Preview

Duplicate

Split (1)

train

The full dataset viewer is not available (click to read why). Only showing a preview of the rows.

The dataset generation failed because of a cast error

Error code:   DatasetGenerationCastError
Exception:    DatasetGenerationCastError
Message:      An error occurred while generating the dataset

All the data files must have the same columns, but at some point there are 3 new columns ({'strategy', 'cumulative_degradation_exposure', 'n_retrains'}) and 5 missing columns ({'regime', 'floor', 'ensemble_mae', 'naive_mae', 'improvement_pct'}).

This happened while the csv dataset builder was generating data using

hf://datasets/the-piyushgoel/drift-forecast-results/deployment_ember/phase4_ember_results.csv (at revision 5641e8a5e365007c33076c2b75c518a9b826def7), ['hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/ablations/ablation_naive_floor_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/deployment_ember/phase4_ember_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/deployment_ember/phase4_ember_sensitivity.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/airlines_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/airlines_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/airlines_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/covtype_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/covtype_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/covtype_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/creditcard_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/creditcard_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/creditcard_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/elec2_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/elec2_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/elec2_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_abrupt_balanced_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_abrupt_balanced_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_abrupt_balanced_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_gradual_balanced_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_gradual_balanced_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_gradual_balanced_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_incremental_balanced_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_incremental_balanced_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_incremental_balanced_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/maliciousurl_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/maliciousurl_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/maliciousurl_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/phase2_ember_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/phishing_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/phishing_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/phishing_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/synthetic/phase1_local_hyperplane_results_raw.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/synthetic/phase1_local_hyperplane_summary_stats.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/synthetic/phase1_local_results_raw.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/synthetic/phase1_local_summary_stats.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/synthetic/phase3_local_results.csv']

Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1848, in _prepare_split_single
                  writer.write_table(table)
                  ~~~~~~~~~~~~~~~~~~^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 784, in write_table
                  self._write_table(pa_table, writer_batch_size=writer_batch_size)
                  ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 795, in _write_table
                  pa_table = table_cast(pa_table, self._schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              strategy: string
              n_retrains: int64
              cumulative_degradation_exposure: double
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 673
              to
              {'regime': Value('string'), 'floor': Value('float64'), 'ensemble_mae': Value('float64'), 'naive_mae': Value('float64'), 'improvement_pct': Value('float64')}
              because column names don't match
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1850, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
                  ...<4 lines>...
                  )
              datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
              
              All the data files must have the same columns, but at some point there are 3 new columns ({'strategy', 'cumulative_degradation_exposure', 'n_retrains'}) and 5 missing columns ({'regime', 'floor', 'ensemble_mae', 'naive_mae', 'improvement_pct'}).
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/the-piyushgoel/drift-forecast-results/deployment_ember/phase4_ember_results.csv (at revision 5641e8a5e365007c33076c2b75c518a9b826def7), ['hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/ablations/ablation_naive_floor_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/deployment_ember/phase4_ember_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/deployment_ember/phase4_ember_sensitivity.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/airlines_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/airlines_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/airlines_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/covtype_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/covtype_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/covtype_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/creditcard_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/creditcard_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/creditcard_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/elec2_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/elec2_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/elec2_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_abrupt_balanced_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_abrupt_balanced_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_abrupt_balanced_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_gradual_balanced_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_gradual_balanced_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_gradual_balanced_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_incremental_balanced_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_incremental_balanced_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/insects_incremental_balanced_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/maliciousurl_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/maliciousurl_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/maliciousurl_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/phase2_ember_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/phishing_phase1_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/phishing_phase2_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/real_world/phishing_phase4_results.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/synthetic/phase1_local_hyperplane_results_raw.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/synthetic/phase1_local_hyperplane_summary_stats.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/synthetic/phase1_local_results_raw.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/synthetic/phase1_local_summary_stats.csv', 'hf://datasets/the-piyushgoel/drift-forecast-results@5641e8a5e365007c33076c2b75c518a9b826def7/synthetic/phase3_local_results.csv']
              
              Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

regime string	floor float64	ensemble_mae float64	naive_mae float64	improvement_pct float64
Hyperplane/exponential (hardest regime)	0	0.688273	0.710946	3.189068
Hyperplane/exponential (hardest regime)	0.05	0.689154	0.710946	3.06511
Hyperplane/exponential (hardest regime)	0.1	0.690048	0.710946	2.939363
Hyperplane/exponential (hardest regime)	0.2	0.691906	0.710946	2.678046
Hyperplane/exponential (hardest regime)	0.3	0.693851	0.710946	2.404536
SEA/linear (easy regime)	0	0.286186	0.458981	37.647657
SEA/linear (easy regime)	0.05	0.288052	0.458981	37.240995
SEA/linear (easy regime)	0.1	0.290926	0.458981	36.614938
SEA/linear (easy regime)	0.2	0.299565	0.458981	34.732538
SEA/linear (easy regime)	0.3	0.311932	0.458981	32.038115
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	0.052301	0.053373	null
null	null	0.034873	0.035553	null
null	null	0.040164	0.048646	null
null	null	0.045835	0.048623	null
null	null	0.044868	0.050114	null
null	null	0.059628	0.078836	null
null	null	0.04655	0.074698	null
null	null	0.05267	0.054052	null
null	null	0.05986	0.069704	null
null	null	0.0462	0.096869	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	0.08658	0.08658	null
null	null	0.114865	0.114865	null
null	null	0.129447	0.129447	null
null	null	0.080436	0.086983	null
null	null	0.063379	0.076987	null
null	null	0.040846	0.070625	null
null	null	0.046605	0.050405	null
null	null	0.060958	0.08973	null
null	null	0.069696	0.063458	null
null	null	0.07645	0.074054	null
null	null	0.081564	0.110549	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	null	null	null
null	null	0.003006	0.003006	null

End of preview.

Forecast-Aware Degradation Attribution: Experiment Results

Result files for the manuscript "Forecast-Aware Degradation Attribution: A Regime-Aware Ensemble and Ratio-Based Diagnostic for Distinguishing Feature Drift from Concept Drift Before It Happens" (in preparation).

Authors

Department of Computer Science, Krishna Institute of Engineering & Technology (KIET), Ghaziabad, Uttar Pradesh, India

Piyush Goel* (corresponding author) - piyush.goel2704@gmail.com
Payal Goswami
Premansh Tripathi
Yukta Arora
Akash Goel (faculty mentor)

*Corresponding author

What this repository contains

Only the generated result files (CSV) from our experiments. The raw real-world datasets are NOT redistributed here. Please obtain them from their original sources (see "Source datasets") and use the loaders in the GitHub repository.

Folder structure

Folder	Contents	Paper section
`synthetic/`	Forecasting results on the SEA and Hyperplane generators (raw and summary stats) and root-cause classification results	Tables II, VII; root-cause classification
`real_world/`	Per-dataset results for 10 real-world datasets: `_phase1_` (forecasting), `_phase2_` (attribution vs. degradation), `_phase4_` (deployment strategy comparison)	Tables III, V, VI
`deployment_ember/`	EMBER2018 deployment comparison and its threshold / label-delay sensitivity analysis	Table IV
`ablations/`	Naive-persistence safety floor sensitivity	Table VIII

File naming

phase1: drift-metric forecasting
phase2: attribution score vs. realized accuracy drop (calibration)
phase3: root-cause classification (synthetic, known drift type)
phase4: deployment strategy comparison (reactive / periodic / proactive)

Source datasets (not included)

EMBER2018, ELEC2, Airlines, Forest CoverType, INSECTS (abrupt, gradual, incremental; balanced variants), CreditCard fraud, Phishing, MaliciousURL. Each dataset is subject to its own license and terms; please consult the original source before use.

Reproducing the results

See the README of the GitHub repository for environment setup (requirements.txt) and run instructions.

Citation

To be updated after publication.

Notes

Results were produced with the real forecasting models (ARIMA, XGBoost), SHAP and ADWIN described in the paper.
Several results are mixed or negative, and are reported as such in the paper.

Contact

Piyush Goel - piyush.goel2704@gmail.com

Downloads last month: 60

Total file size:

9.69 MB