Partially ordered rough ensemble clustering for multigranular representations

Lingras, Pawan; Haider, Farhana

doi:10.3233/ida-150772

Cited by 4 publications

(1 citation statement)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the clustering attribute selection based on rough cluster analysis, the attribute is selected to split according to the roughness of the attribute or the attribute dependencies, e.g., Min-Min Roughness (MMR) [43], Maximum Dependency of Attribute (MDA) [44], Maximum Attribute Relative (MAR) [45], and Mean of Accuracy of approximation using variable precision of attributes (MA) [46], all of which are related to approximations computation. In rough set based cluster ensemble method [47,48,49], the computation of approximations is a necessary step. Hence, it is meaningful and valuable to develop an efficient algorithm for computing approximations, which can accelerate the data mining or machining learning related tasks under a dynamic environment, e.g., attribute reduction, rule induction, cluster analysis and cluster ensemble.…”

Section: Introductionmentioning

confidence: 99%

Incremental fuzzy probabilistic rough sets over two universes

Luo

et al. 2017

International Journal of Approximate Reasoning

View full text Add to dashboard Cite

The fuzzy Information System over Two Universes (ISTU) formalizing a data table corresponding to two universes as well as their relations is common in real-world applications, e.g., recommender system and clinical diagnosis system. The fuzzy probabilistic rough sets over two universes (FPRSMTU) can deal with a fuzzy relation and allow a tolerance inaccuracy in the construction of rough approximations in a fuzzy ISTU, which is a generalization of classic rough sets with fuzzy and probabilistic theories. As a necessary step for knowledge discovery based on rough sets, the fuzzy rough approximations of fuzzy ISTU need to be updated efficiently under dynamic data environment. Incremental technique is an efficient approach for dynamic information processing by making full use of previously obtained knowledge. In this paper, incremental approaches for updating approximations of fuzzy ISTU are proposed while some objects adding into or deleting from the two universes, and the corresponding incremental algorithms are designed. Experimental evaluations on real datasets as well as artificial datasets show the effectiveness of the proposed incremental updating method compared with the non-incremental method.

show abstract