The present embodiment is able to find the optimal or near optimal variables composition of multivariate models by an evolutionary process within acceptable amount of time and resources that are less than using full variables permutation methodology. Subjected to any data, it adaptively identifies and constructs the most effective combination of the relevant variables to achieve one or more objectives. The objective could be for high explanatory power, high predictive power, response measure, or other objectives that the user defines. The present embodiment solves the sequential F-test problem by conducting non-sequential and non-linear search. The algorithm also solves partial F-test dilemma by evaluating all candidate variables membership intact, maintaining fidelity of full variables membership test throughout its permutation. Furthermore, the stochastic nature of the algorithm neutralizes the prejudices of manual decisions in variables identification and membership construction.