摘要
针对BP_Adaboost算法预测精度不高的问题,对算法作了改进:先用遗传算法对每个BP神经网络弱预测模型进行优化;然后把优化后的BP神经网络模型看作为新的弱预测器;再通过BP_Adaboost算法,用多个被遗传算法优化后的BP神经网络弱预测器组成强预测器模型;最后加权整合优化后用2 000组随机数据验证改进后算法的预测精度,用Matlab程序仿真实现改进后的BP_Adaboost算法,并与改进前的BP_Adaboost算法作比较。程序运行结果表明,改进后的BP_Adaboost算法预测精度有了明显提高。
The prediction problem is the core of large data.The existing BP_Adaboost algorithm is a fusion in BP neural network model of prediction model algorithm.As the accuracy of BP_Adaboost algorithm is not high, the BP_Adaboost algorithm is improved in our stndy.BP neural network model is put as a weak predictor,and the strong predictor of multiple BP neural network composed of weak predictor is obtained by BP_Adaboost algo-rithm.Genetic algorithm is used for each BP neural network prediction model for optimization.When optimized BP neural network model as a new weak predictor,and through the BP_Adaboost algorithm,the BP neural net-work by genetic algorithm optimization of weak predictor is composed of strong predictor model.From 2 000 groups of random experimental data,the prediction accuracy to verify the improved algorithm leads to improved BP_Adaboost algorithm simulation with Matlab program.The result is compared with the BP_Adaboost algo-rithm before improvement.From the result of running program,the prediction of the improved BP_Adaboost al-gorithm possesses higher precision.
出处
《桂林理工大学学报》
CAS
北大核心
2014年第3期589-594,共6页
Journal of Guilin University of Technology
