Adaptive simulated annealing genetic algorithm used for sleep electroencephalogram staging feature selection

Abstract

The invention discloses an adaptive simulated annealing genetic algorithm used for sleep electroencephalogram staging feature selection. Sleep staging is performed through electroencephalogram signals, a large number of feature parameters require to be extracted out of the electroencephalogram signals, and the relatively optimal feature parameter combination is selected out through screening to be used for establishing a sleep electroencephalogram mathematical model. In the present simulated annealing genetic algorithm, the high overall search capacity of the genetic algorithm and the high local search capacity of the simulated annealing algorithm are reserved so as to enhance the probability of generating excellent individuals. In the simulated annealing operation of the present algorithm performed on the individuals in the iterative process, the mechanism for randomly generating new solutions in the neighborhood of the current optimal solution has the fatal flaw. The algorithm aims at the flaw and solves the disadvantages that the neighborhood new solution generation mechanism of the conventional simulated annealing genetic algorithm has low iterative efficiency and is greatly affected by the neighborhood range and can realize adaptive adjustment of crossover probability and mutation probability, and the fitness function can be designed by using the weighing method.

Description

technical field [0001] The invention relates to an adaptive simulated annealing genetic algorithm for feature selection of sleep EEG staging, and belongs to the technical field of adaptive simulated annealing genetic algorithm. Background technique [0002] Feature selection is to select a small number of feature parameters that are most suitable for different tasks from all feature parameter sets through an algorithm. When the number of characteristic parameters is small, the optimal characteristic parameters can be screened by calculating all the combination of characteristic parameters by exhaustive method. When the number of characteristic parameters is large, the exhaustive method is almost impossible to implement. With the development of pattern recognition algorithms, feature parameter selection has become an important means of optimizing feature parameter classification models. Feature parameter selection can not only simplify the established feature classification...

AI Technical Summary

Problems solved by technology

However, the shortcomings of this algorithm are: (1) It is necessary to design the type and parameters of the fitness function according to the actual optimization problem, and there is no general fitness function standard; (2) it is easy to fall into local optimum; (3) the convergence speed slow

However, when the algorithm performs simulated annealing operations on individuals in the iterative process, the mechanism of randomly generating new

Images