论文部分内容阅读
针对机场噪声的预测问题提出两种先模糊聚类再支持向量回归的时间序列预测方法.一种是模糊C均值聚类再回归,通过聚类将同簇样本限定在一定区域内,然后对同簇样本进行回归预测.另一种是基于阴影集的粗糙模糊C均值聚类再回归,通过聚类将簇划分为核心区和边界区,属于核心区的样本对簇的贡献比属于边界区的样本大,将样本限定在同簇同一区域的范围内,再对同簇相似样本进行回归预测.选用两个常用数据集和北京某机场实测数据进行实验.结果表明,基于模糊聚类的先聚类再回归方法比直接回归方法得到的拟合值更精确.
Aiming at the problem of airport noise prediction, two time series forecasting methods based on fuzzy clustering and then support vector regression are proposed. One is fuzzy C-means clustering and regressing, and cluster samples are used to limit the same-cluster samples to a certain region, The other is based on the shadow set of rough fuzzy C-means clustering regression, the cluster is divided into clusters by the cluster core area and the boundary area, belonging to the core area of the sample contribution to the cluster than belonging to the border area The sample is large, the sample is limited to the same region of the same cluster, and then regression prediction is made on the same cluster of similar samples.Experimental results of two commonly used datasets and measured data from an airport in Beijing show that the first clustering based on fuzzy clustering The class regression method is more accurate than the direct regression method.