Saturday, September 15, 2007

PCA and exploratory analysis for linkage learning in Genetic algorithm

I just came up with the idea of applying Principle component analysis (PCA) and other feature selection (correlation) analysis tools in data mining to the linkage detection in genetic algorithms based optimization.

During the evolutionary process, we can calculate the correlation( mutual information) of the features with respect to the fitness values, which can be used to guide the new individual generation process.

George Hu

No comments: