RCGNN:一种基于实数编码的遗传神经网络预测方法

RCGNN:A Real Number Coding Genetic Neural Network Prediction Model

将遗传算法与神经网络相结合,提出一种实数编码、自适应选择、算术交叉、高斯变异、爬山操作的改进遗传BP神经网络RCGNN,利用遗传算法对神经网络权值和阈值进行优化。以时间序列预测的实例进行编程计算表明,用遗传算法进行网络训练,其收敛速度快,最终总误差最小,预测准确率高。对算法中参数进行的相应研究表明,增加爬山操作次数能很好地提高网络训练的速度,同时使误差下降快;爬山操作越多,收敛速度越快,最终误差越小,但计算运行时间也会增加。

In order to overcome the shortcomings of traditional error back propagation algorithm for updating the weights of the multi-layer forward neural networks, such as the low precision of the solutions, the slow search speed and easy convergence to the local minimum points, we propose an improved genetic neural network RCGNN, which combines genetic algorithm and neural network, contains real number coding, arithmetic crossover, Gauss mutation and climbing operation. The result of a illustrative sequence prediction example shows that, the method has a faster training speed, less total error and can predict more accurate. When adjusting the parameters in the algorithm, we discover that adding climbing operation times can effectively quicken the training speed of network and lessen the error diminished faster,more climbing operation can get quicker convergence speed and smaller final total error, but need more time to run this process.