摘要
基于粗集、遗传神经网络的环境质量评价方法利用粗集对属性的归约功能将数据库中的数据进行归约,并将归约后的数据作为训练数据提供给BP神经网络;再用遗传算法和BP算法相结合的混合算法来训练网络预测模型的结构(在得到最优网络结构的同时也得到网络的最优权值和阈值)。通过粗糙集归约,提高了训练数据表达的清晰度,也减小了BP神经网络的规模,同时利用BP神经网络又克服了粗糙集对噪声数据敏感的影响。这一算法克服了BP算法收敛速度慢、易陷入局部极小等缺陷,实例证明提高了预测精度。
The method of environmental quality assessment which was based on rough sets and genetic-neural network reduced datum from database by using rough sets reduction function, and then transferred the reduced datum to the BP neural network as training datum. The genetic algorithm, combined with the BP neural network, trained the structure of the network prediction model （received the optimum weights, threshold values and the optimum structure） .By datum reduction, the expression of training would become clear, and the scale of neural network would be simplified. At the same time, neural network solved rough sets problem of yawp sensitivity.It overcome the limitations of the back-propagation algorithm in slow convergent rate and got into local optima.The example demonstrated that this method improved the prediction precision.
出处
《云南地理环境研究》
2006年第1期97-100,共4页
Yunnan Geographic Environment Research
关键词
环境质量评价
属性归约
遗传神经网络
混合算法
预测精度
environmental quality assessment
datum reduction
genetic-neural network
the combined algorithm
the prediction precision
