Baiyang Chen, Yongxiang Li, Dezhong Peng, Hongmei Chen, and Zhong Yuan*, "Fusing multi-scale fuzzy information to detect outliers," (PDF) Information Fusion, vol. 103, p. 102133, 8 November 2023, DOI: 10.1016/j.inffus.2023.102133
Outlier detection aims to find objects that behave differently from the majority of the data. Existing unsupervised approaches often process data with a single scale, which may not capture the multi-scale nature of the data. In this paper, we propose a novel information fusion model based on multi-scale fuzzy granules and an unsupervised outlier detection algorithm with the fuzzy rough set theory. First, a multi-scale information fusion model is formulated based on fuzzy granules. Then we employ fuzzy approximations to define the outlier factor of multi-scale fuzzy granules centered at each data point. Finally, the outlier score is calculated by aggregating the outlier factors of a set of multi-scale fuzzy granules.
- python=3.8
- numpy=1.23
- scikit-learn=1.2
To reproduce the results in the paper:
python run_reproduce.py
To reproduce the examples in the paper:
python run_example.py
To run MFIOD on customized datastes with default parameters:
# Assume the dataset be saved in a Numpy npz file with n samples and m dimensions
# An m dimensional bool vector be given to indicate: True=Nominal attribute, False=Numerical attribute; if not provided, all attributes are treated as numerical.
python run_customs_default_paras.py
To run MFIOD on customized datastes with parameter tuning:
To be updated later.
If you find the code or datasets useful in your research, please consider citing:
@article{Chen2024MFIOD,
title = {Fusing multi-scale fuzzy information to detect outliers},
author = {Baiyang Chen and Yongxiang Li and Dezhong Peng and Hongmei Chen and Zhong Yuan},
journal = {Information Fusion},
volume = {103},
pages = {102133},
year = {2024},
issn = {1566-2535},
doi = {10.1016/j.inffus.2023.102133},
}
or:
Baiyang Chen, Yongxiang Li, Dezhong Peng, Hongmei Chen, and Zhong Yuan, "Fusing multi-scale fuzzy information to detect outliers," Information Fusion, vol. 103, p. 102133, doi: 10.1016/j.inffus.2023.102133
If you have any questions, please contact farstars@qq.com.