Optimization of Random Forests Algorithm Based on ReliefF-SA

Abstract

As a widely used classification integration algorithm, the random forests based on decision tree combination is mostly used for classification and regression problems, but its deficiencies still need to be improved. In this paper, we propose simulated annealing algorithm, and its parameters is combined with the global search optimal solution and local optimal solution to optimize search. Then, we propose a random forests hybrid algorithm by combining the Relief algorithm and annealing algorithm, briefly introduce its thoughts and algorithm flow, and also carry out simulation experiments. According to the experimental results, the classification effect of the hybrid algorithm is generally the best, which greatly improves the overall generalization performance of the random forests.