小波神经网络遗传算法及其在矿山压力预报中的应用

GENETIC ALGORITHM OF WAVELET NETWORK AND ITS APPLICATION IN FORECASTING GROUND PRESSURE

为了克服ＢＰ网络自身算法的缺陷，得到更高的学习精度和更快的收敛速度，将遗传算法、小波分析、人工神经网络和模拟退火思想结合起来，提出了一种遗传小波网络：即用遗传算法来学习小波神经网络层间的权值、尺度参数和位置参数。用两维和三维ＸＯＲ问题对其性能分别进行了测试，取得了理想的效果；将其应用于矿压预报，得到了比传统神经网络更优的效果。

In order to overcome the algorithm shortcoming of BP network and obtain much higher accuracy and faster speed, a wavelet network based on genetic algorithm was put forward, which combined genetic algorithm with wavelet analysis and artificial neural network and simulated annealing thought, to study various parameters such as weights etc by genetic algorithm. Its property was tested by solving the problem of twodimension XOR and threedimension XOR, and good results of them were obtained. And it was also used to solve the problem of forecast in the field of ground pressure, more optimal results than typical BP network have been obtained.