基于 PSO－SVM算法的环境监测数据异常检测和缺失补全

摘要

For problems of abnormal data and missing data in environmental monitoring, an anomaly detec-tion and data missing completion algorithm was presented based on particle swarm optimization with support vec-tor machine ( PSO-SVM) .Non-linear SVM model was established by applying the PSO algorithm in selecting the appropriate training parameter set and fitting prediction of real data.Taking the experimental data from a sewage plant in Ningxia Hui Autonomous Region, the predictions by this algorithm had the accuracy rate of 97.977%, showing high accuracy in abnormal data detection and missing data completion.%针对环境监测数据异常和数据缺失问题，提出了基于支持向量机的粒子群优化数据异常检测和缺失补全算法。利用粒子群优化算法选取较优的支持向量机训练参数组合，以此建立非线性的支持向量机模型，并利用结果模型对测得的真实数据拟合预测。以宁夏回族自治区某污水处理厂的污染物测量数据作为实验数据，结果表明，利用该算法预测数据的准确率可达97．977％，检测异常数据准确度高，缺失数据补全正确。