HIPE Feature Data Set
This is the companion website for the publication
Michael Vollmer, Holger Trittenbach, Shahab Karrari, Adrian Englhardt, Pawel Bielski, Klemens Böhm, “Energy Time-Series Features for Emerging Applications on the Basis of Human-Readable Machine Descriptions” accepted at Second International Workshop on Energy Data and Analytics (ACM e-Energy Workshop 2019), 05 Mar 2019.
This website provides the feature data set used in the manuscript to download, and code to reproduce the feature extraction.
Resources
The current version of the HIPE Feature Data Set 1.0.0 (2019-03-05) is based on the 3 month (2017-10-01 to 2018-01-01) HIPE data set. We provide four different feature representations of the HIPE data set. Each data set contains 36 features for 7 electrical attributes of 11 machines.
- No filtering, 15 minute aggregation: Features-15Min-Full (49.2 MB packed)
- No filtering, 1 hour aggregation: Features-1H-Full (14.0 MB packed)
- Machine on filtering, 15 minute aggregation: Features-15Min-On (22.8 MB packed)
- Machine on filtering, 1 hour aggregation: Features-1H-On (6.7 MB packed)
The archives contain one file for each machine. The CSV format of each file is id,weekday,ATTRIBUTE1_FEATURE1,ATTRIBUTE1_FEATURE2... where id is the timestamp and weekday starts with Monday being value 1. The feature extraction code to reproduce the data set is available on Github. For further details and experimental results, we refer to the companion paper.
The code is licensed under a MIT License and the data set under a Creative Commons Attribution 4.0 International License.
Contact
For questions and comments, please contact michael.vollmer∂kit.edu