• 关键词：水质指标  预测模型  关联向量机回归  支持向量机回归

• 基金项目：国家自然科学基金(No.51305407)

• 笪英云
• 浙江师范大学数理与信息工程学院, 金华 321004

• 汪晓东
• 浙江师范大学数理与信息工程学院, 金华 321004

• 赵永刚
• 江苏省环境监测中心, 南京 210036

• 蒋敏兰
• 浙江师范大学数理与信息工程学院, 金华 321004

• 叶美盈
• 浙江师范大学数理与信息工程学院, 金华 321004

• 摘要：提出了一种基于关联向量机回归的水质时间序列预测模型,并以该模型对氢离子浓度指数(pH值)、溶解氧(DO)、高锰酸盐指数(COD_Mn)和氨氮(NH₃-N) 4种重要水质指标进行预测.首先采用国家环保部发布的四川攀枝花龙洞水质自动监测数据进行实验,对该模型的有效性进行了验证;然后将关联向量机回归预测模型与支持向量机回归预测模型进行比较.为了比较不同核函数的预测效果,实验中预测模型的核函数分别采用了线性函数和高斯函数.实验结果表明,关联向量机回归模型的预测效果不亚于支持向量机回归模型;且在给出预测值时,还能同时给出预测结果的可信程度.

• Abstract：Water quality time series prediction model based on a relevance vector machine regression is proposed, which is applied to predict the four important water quality indexes of pH value, DO, COD_Mn and NH₃-N. The Panzhihua Cave automatic water quality monitoring data released by the Ministry of Environmental is used to validate the effectiveness of the relevance vector machine regression prediction model, and then compared with the support vector machine regression prediction model. The linear and gaussian functions are selected as the kernel function to contrast the result of different kernel functions. Experimental results indicate that the relevance vector machine regression prediction model is no less than the support vector machine regression prediction model. Furthermore, the relevance vector machine regression can give the prediction values and also compute the confidence levels for prediction results.

• Key words：water quality index  prediction model  relevance vector machine regression  support vector machine regression

• 
  摘要点击次数： 4493 全文下载次数： 2194