基于模糊神经网络的数据挖掘算法

A Data Mining Algorithm Based On Fuzzy Neural Net

提出一种基于模糊神经网络进行数据挖掘的新方法。构成模糊神经网络的模糊化层采用高斯函数计算5个模糊隶属度,高斯函数需要的均值、方差以及隶属度的中心值都可通过预先计算采集到的数据得到。模糊推理层采用取小取大运算代替常用的积和运算,加快了网络的推理速度。在模糊神经网络训练阶段,首先利用重心法对模糊化层输出进行反模糊化,再采用BP思想,利用梯度法求误差值并进行反传调整隶属度函数的参数值。为提高网络推理精度和速度,通过设立相应的规则对网络进行裁剪,剪掉多余的节点和权值,最后依据一定的思想对产生的模糊规则进行简化和提取。以工业锻造中的智能温度控制系统的控制数据为例进行仿真,结果表明,该网络具有较高的精度和较好的鲁棒性。

An algorithm of Data Mining based on fuzzy neural net is proposed in the article. The algorithm creates a new fuzzy neural net, in which Gaussian - Function is used to calculate five fuzzy memberships in fuzzy layer. The average and variance required by Gaussian - Function can be calculated in advance through the data to be collected. Using of fuzzy inference from a Max - min operation to replace commonly multiply - add operation accelerates the speed of the network. In training stage, first the centre - of - gravity method is used to resist the fuzzy of the output of the fuzzy layer, and then BP idea is adopted to calculate the error and adjust the membership function parameters. To improve the accuracy and speed of fuzzy neural network, the net, using appropriate rules, crops redundant nodes and weights to extract and simplify the rules. A simulation is performed by using the control data from the intelligent temperature control system in industrial forging, the result shows that the algorithm has higher precision and stronger robustness.