模糊推理与神经网络、遗传算法在系统预测中的应用

近年来，原来并没有什么必然联系的模糊理论和人工神经网络有了较多的结合．这是由于在模糊系统设计涉及到的3个问题：第一，模糊规则的选取}第二，模糊概率函数（隶属函数）的确定；第三，模糊决策算法的决定．这在有些系统中并不是十分明确的，而人工神经网络却不需要人为干预，而只需通过实际输入、输出数据的学习即可得到其决策．但是人工神经网络也有其缺点，它不够稳定，单独使用确定度不是很高，而且神经网络中所使用的BP算法有其限制性，只能得到局部最优解．因此，我们将模糊理论与神经网络相结合，根据其预测值与实际值的误差选取适当的权重，建立二者对预测结果的影响关系的模型，用于系统预测．在神经网络中，把BP算法和遗传算法结合起来，得到全局最优解．

The Application of Fuzzy Inference and Neural Network, Genetic Algorithm in the System Prediction

In recent years, the fuzzy theory and artificial neural network which used to have no relation has had much connection. There are three questions related to the design of fuzzy system; firstly, the selection of fuzzy regulations; secondly, the determination of fuzzy probability function ( subordination function ) ; thirdly, the consideration of fuzzy decision. They are not very clear in some systems, but artificial neural network don't need any artificial interference, and what it needs to acquire decision is only learning actual input and output. However, artificial neural network is not sufficiently stable, and the creditability is not satisfactory, and only partial optimal resolution can be obtained. Consequently, combining the fuzzy theory with the neural network, we choose proper weight according to the proportion of the prediction error. Then we can build a model to be applied to the systematic prediction. Meanwhile, we try to combine BP with GA in the neural network to gain the total resolution.