Research On Hybrid Learning And Ensemble Learning Based On Genetic Algorithm

In the field of intelligence learning, hybrid intelligence learning has been the research hot recently. Encouraging achievements and extensive applications have been made in many fields. Hybrid methods which combine three or more intelligence technologies are developed to solve the problems of complex systems. Simultaneously, another learning method——ensemble learning, which has obtained good predictive results in application, draws more attention. This paper has finished some research work and experiments with the aid of the two methods——hybrid learning and ensemble learning.Based on the study of hybrid learning and ensemble learning, two algorithms have been developed to solve the combined optimization problem and the parameters-choosing problem respectively. And this paper has employed the genetic algorithms to evolve the weights of ensemble learning so as to improve the efficiency of ensemble learning.The main work of this paper is:(1) Based on hybrid learning, a hybrid algorithm which combines the partheno-genetic algorithm and simulated annealing algorithm has been developed to solve the combined optimization problem. The simulating results of the vehicle routing problem have proved the efficiency of the hybrid algorithm.(2) In order to deal with the parametric-choosing problem, many genetic algorithms with different crossover and mutation probabilities have been employed to construct ensemble algorithm GAE. Analysis of the learning results has also given in the paper.(3) Boosting and Bagging which are the representative algorithms of ensemble learning have been applied to neural network, and encouraging results have been obtained. This paper also uses genetic algorithm to adjust the weights of single algorithms in ensemble learning, Simulation has been made based on UCI database set.