All subsets regression using a genetic algorithm

Subset regression procedures have been shown to provide better overall performance than stepwise regression procedures. However it is difficult to use them when a large number of candidate variables exists due to the high computational costs caused by the combinatorial nature of evaluating each potential subset. To resolve this difficulty, the use of Genetic Algorithm search algorithm is proposed to reduce the number of subsets that must be evaluated. We also propose using an information complexity based goodness of fit measure which penalizes over-fitting, a common problem associated with the use of R².