Brute Exhaustive Optimization of Intelligent Small Weighted Voting Ensembles in 1EXP(-)Z+ Initial-Term based Arithmetic Sequence’s Multi Precision Search Spaces

Other than the individual machine learning models' capabilities, the weighted voting ensemble (WVE) technique relies on appropriate weight assignment in order to significantly realize prediction performance improvement.Often evolutionary global or grid local search heuristics are being applied for such a challenging optimization task.However, these techniques do not guarantee optimal solution finding.In turn, the surprising outstanding successes of brute exhaustive search procedure in producing similar results shed light on its significance and the need to exploit its possible weights solutions search space(s) with corresponding sizes as a key determinant factor for implementing a successful brute search procedure for finding optimal WVE solution with a trade-off the computational efficiency.This paper formulates an asymptotically WVE weights domain constraints optimal 1EX(-)Z + initial term-based arithmetic sequences initialization function, and then a computational multi-precision search space-based generation algorithm is developed to find optimal WVE solution as part of the brute exhaustive search procedure.It took 45 minutes for a proposed algorithm to generate 133,192 combinations and find the optimal solution in weights space of precision 0.01.