神经网络的分类、聚类功能及其规则抽取研究

【作者】 雷景生；

【导师】 孟吉翔；

【作者基本信息】 新疆大学 ， 应用数学， 2003， 博士

【摘要】 基于数据挖掘的观点，本文在前人研究的基础上，对神经网络的分类、聚类功能及其规则抽取进行研究，取得了一些有价值的结果。 在神经网络分类器研究中，1)提出了一种确定BP网络分类器初值的方法，用来提高BP网络的收敛速度，降低分类误差，避免局部极小问题。2)在FTART2的基础上，提出了一种快速自适应神经网络分类器FANNC。同时，对FANNC的容错性网络训练提出了一种有效的方法。3)提出了一种新的构造神经网络分类器集成的方法，理论分析及实验结果表明，其分类效果较好。 在神经网络的聚类功能研究中，1)在分析SOM算法存在问题的基础上，提出了一种模糊自组织映照网络(FSOM)算法，克服了传统SOM算法的不足。2)分析了基于LVQ的FCNN算法的局限性，并对其进行改进，结果表明，改进后的学习算法可以有效克服原有算法的不足并具有较高的收敛速度。 在神经网络规则抽取方法研究中，提出了一种从FANNC网络中抽取if-then规则的方法，试验结果表明，该方法能抽取可理解性好、预测精度高的if-then规则。更多还原

【Abstract】 Point of view base on data mining, the foundation on many scholars to research, the thesis study of classification , clustering funtion and rule extraction in Neural Networks, many counts results acquired. Study of classifier in Neural Networks, 1) This paper presents fix an initializing algorithm on classifier in BP Neural Networks. The algorithm can not only speed up convergence of BP neural networks and reduce the error of training, but also abstain abstain from converging to local minimum point. 2) The foundation on FTART2, this paper presents a fast adaptive neural network classifier (FANNC), at the same time, an adaptive fault-tolerant neural networks learning algorithm is proposed. 3) A new approach to classifier of neural network ensemble is proposed. Theoretical analyses and experimental results show that this approach outperforms the traditional ones that ensemble all of the individual networks.Study of clustering funtion in Neural Networks, 1) the foundation on analyses exist ’ problem of SOM algorithm, this paper presents an algorithm of the fuzzy self-organizing feature map (FSOM), tide over shortcoming of the traditional SOM algorithm. 2) Demonstrated by analysis are weaknesses of the current learning algorithm of fuzzy clustering neural networks FCNN, and the algorithm is improved. The result of simulation verifies the proposal properly.Study of rule extraction in Neural Networks, we proposed a method that is able to extract if-then ruler from trained FANNC networks in this paper, Experimental result shows that those if-then rules are comprehensible, accurate.更多还原