| Feature selection method |
Author Name and year of publication |
Description |
Drawback |
| Feature and Relation Selection (FARS) |
B. Hu, H. Liu, J. He and X. Du (2008) |
Evaluated by using table symmetrical uncertainty (TSU) which is symmetrical uncertainty (SU) value between relation and class. (over multi-relational dataset) |
Discrete values are not handled |
| Feature Selection using InfoDist |
C. Sha, X. Qiu and A. Zhou (2007) |
Evaluated by InfoDist which based on information theory. |
Discrete values are not handled |
| Wilks Lambda criterion method |
A. Ouardighi, A. Akadi and D. Aboutajdine (2007) |
Evaluated by a statistical value used in discriminant analysis. |
Insufficient to improve the classifier performances only by relevance criterion. |
| MR2 feature selection method |
A Unler, A Murat, RB Chinnam (2007) |
This method uses InfoDist and Pearson Correlation to calculate the relevant features (over multi-relational dataset) |
Cutoff value is hard to decide |
| Fast Correlation Based Filter (FCBF) |
L. Yu and H. Liu(2003) |
Evaluated by information gain combines optimal subset and feature relevance weight method. |
Discrete values are not handled |
