针对传统时间序列预测融合算法对于具有非线性、随机性和突发性的数据拟合度不佳的问题,提出了一种基于灰色最小二乘支持向量机(GM-LSSVM)预测的时序数据融合方法。利用少量监测数据对模型进行训练,以灰色回归预测数据作为最小二乘支持向量机的输入数据,并对下一步未知信息进行预测,以达到减少通信开销的目的。实际测量结果表明,该方法所需样本数量较少,预测准确率较高,能有效降低数据传输开销。
The traditional approaches for temporal data aggregation do not take the nonlinear,random and mutation of time series data into account.For this problem,an aggregation method based on GM-LSSVM combination prediction is proposed.In this method,grey model(GM) prediction theory is introduced into least squares support vector machines(LSSVM).A real-world data set and a random nonlinear data set from UCI are used to verify the method.The results show that it is effective to predict the temporal data series and reduce the number of transmissions in wireless sensor networks.