摘要
河流水质实时评价技术对当前河流水资源管理和保护具有重要意义。该文以淮河水质为例,利用粒子群优化的极限学习机(Particle Swarm Optimization-Extreme Learning Machine,PSO-ELM)分类算法对淮河水质进行类别判定。在极限学习机(ELM)分类算法中随机给定输入权值矩阵和隐含层偏置,需要较多的隐含层节点才能达到所需的精度要求,隐含层节点过多易于出现过拟合现象并增加算法的计算量。该文利用粒子群算法(PSO)优化极限学习机的输入权值矩阵和隐含层偏置,计算输出权值矩阵,以减少隐含层节点。通过对比PSO-ELM、ELM这2种算法发现,PSO-ELM算法以较少的隐含层节点可获得更高的精度,降低了对实验样本的需求量,提高了模型的拟合能力。实验结果表明,PSO-ELM对于水质类别判定具有一定的可行性和有效性。
Real time evaluation of rivers＇ water quality has great significance for maintenance and protection of water resource. In the case of Huaihe River, PSO-ELM（Particle Swarm Optimization-Extreme Learning Machine, PSO-ELM） was used to classify the water qualities. In ELM, it will need more hidden layers to achieve the required accuracy when we randomly set up the input weight matrix and hidden layer offset. Excessive hidden layer nodes will be easy to lead to over fitting and increasing the calculated quantities of the algorithm. In order to avoid this problem in ELM, PSO algorithm was utilized to optimize the input weight matrix and the hidden layer offset of ELM, and calculate the output weight matrix to reduce the nodes of hidden layer. Compared with ELM, PSO-ELM can achieve higher accuracy by utilizing lesser nodes in the hidden layer, which can reduce the demanded quantities of experimental samples, and improve the fitting capability of the model. Experimental results show that PSO-ELM has feasibility and effectiveness for judging the water quality classification.
出处
《环境科学与技术》
CAS
CSCD
北大核心
2016年第5期135-139,共5页
Environmental Science & Technology
基金
国家自然科学基金项目(61273068)
上海市自然科学基金项目(12ZR1412600)
上海市教委科研创新项目(13YZ084)
关键词
粒子群优化
极限学习机
水质评价
权值
隐含层
particle swarm optimization
extreme learning machine
water quality assessment
weight
hidden layer
