Application of collaborative filtering methods in the problem of predicting the performance of population optimization algorithms

In this paper we propose an approach to solving the problem of choosing the most efficient algorithm for solving a given continuous optimization problem, based on the using of collaborative filtering methods. A prototype of a software system based on a set of the most popular population optimization algorithms and a system of test objective functions for continuous optimization problems is described. The implementation of several methods for predicting the performance of a given algorithm is considered. The results of computational experiments and comparison of the considered methods are presented.