An unsupervised learning method for novelty detection and characterization based on qualitative spatial relations
Python 3.6+
- GetQSRRelations.py: Generate 4 QSR features to present the state transition, including RCC, QDC, STAR-4, and QTC.
- Existence.py: Generate one additional existence feature.
- Read_Data_consequence.py: Filter state transitions that at least one feature has changed.
- Data_preprocessing.py: Concatenate state traisitions and convert them into clustering-welcomed format.
- Hierarchical_clustering.py: Perform hierarchical clustering on the state transitions.
If you find this method useful, please cite:
@inproceedings{KR2021-43,
title = {{Unsupervised Novelty Characterization in Physical Environments Using Qualitative Spatial Relations}},
author = {Li, Ruiqi and Hua, Hua and Haslum, Patrik and Renz, Jochen},
booktitle = {{Proceedings of the 18th International Conference on Principles of Knowledge Representation and Reasoning}},
pages = {454--464},
year = {2021},
month = {11},
doi = {10.24963/kr.2021/43},
url = {https://doi.org/10.24963/kr.2021/43},
}