Feature selection software, Modeling, data classification and decision making
FEATURE SELECTION SOFTWARE
Our own Soft Computing Software for feature selection and data dimensionality reduction in new products design uses experimentation design. It is suitable for researches focusing on determining relevant features and their relations with the output feature being studied.
This software performs the feature selection by means of a genetic algorithm that obtains a limited set of features based on the following measurements:
- Population Correlation Coefficient (PCC).
- Information Correlation Coefficient (ICC).
- Mutual Information (MI).
The selected input variables are those with higher ICC, PCC and MI values regarding the output.
In order to reduce the information redundancy of the selected features, the mRMR algorithm is implemented. It enables de selection of features with maximum relevance with respect to the output and the minimum redundancy among input variables.
