基于时序相似性的机场噪声监测点交互预测

项目名称： 基于时序相似性的机场噪声监测点交互预测

项目编号： No.61501229

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 无线电电子学、电信技术

项目作者： 陈海燕

作者单位： 南京航空航天大学

项目金额： 19万元

中文摘要： 针对目前机场噪声监测系统可能出现的因监测点异常而无法准确反馈所在区域噪声的问.题，本项目拟建立一种基于监测点之间时序相似性的关联预测模型，用软件的方法来发现监测.点的异常并预测监测失效区域的噪声情况。项目的主要研究内容包括：.1）机场噪声监测点时间序列相似性研究，重点研究监测点之间的时序相似性度量、异常监测点的判定及关联监测点的筛选；.2）基于时间序列相似性的特征加权支持向量机研究，重点研究基于关联监测点集合的预测模型训练集的准备，以及特征加权矩阵的计算和应用；.3）基于加权特征支持向量回归的失效监测点关联预测模型研究，重点研究不同监测频率噪声数据集上，及采用不同特征权值计算方法时，关联预测模型预测能力的验证。

中文关键词： 机场噪声监测；异常检测；交互预测；时间序列相似性；特征加权支持

英文摘要： This project is proposed to solve the problem that if a monitoring point in airport noise monitoring system is failure, noise of a monitoring region can not be accurately feedback. This project intends to establish a interactive prediction model based on time series similarity of the monitoring points, so as to detect the abnormal point and predict the noise of failure region by software way. The main contents include:.1) Research on the time series similarity of the monitoring points, focusing on the timing series similarity measurement, outlier detection and the screening of related monitoring points.;.2) Research on feature weighted support vector machine based on the time series similarity, focusing on the preparation of the training data sets and the calculation and application of feature weights matrix;.3) Research on the interactive prediction model based on the feature weighted support vector regression, focusing on the prediction ability verifications of the interactive prediction models based on the noise data sets of different frequency, and with the feature weights be calculated in different ways.

英文关键词： Airport noise monitoring;Outlier detection ;Interactive prediction;Time series similarity;Feature weighted support vector machine