Abstract
The missing values can be an important obstacle and challenging problem in data analysis. The paper presents the neuro-fuzzy system that handles incomplete data. The system is complete: it can extract the fuzzy rule base from both complete and incomplete data and can elaborate answers for complete and incomplete data. The second major feature of the system is the assignment of weights to attributes in fuzzy rules. The weights are assigned locally: each fuzzy rule has its own weights for attributes. This feature may improve the precision of answers elaborated by the system and may reveal relations between attributes in the data set. The paper is accompanied by experimental results. The results show that the subspace technique is advantageous in handling data set with missing values. The results also reveal that for approximation of complete data it is better to apply techniques without subspace approach.
| Original language | English |
|---|---|
| Pages (from-to) | 30-46 |
| Number of pages | 17 |
| Journal | Fuzzy Sets and Systems |
| Volume | 269 |
| DOIs | |
| Publication status | Published - 15 Jun 2015 |
Keywords
- Incomplete data
- Missing values
- Neuro-fuzzy system
- Rough fuzzy approach
- Subspace analysis
ASJC Scopus subject areas
- Logic
- Artificial Intelligence
Fingerprint
Dive into the research topics of 'Rough subspace neuro-fuzzy system'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver