中文摘要：

综合运用集合经验模态分解（EEMD）和最小二乘支持向量机（LS-SVM）建立了空中交通过点流量预测模型.EEMD的分解结果显示,高频本征模态函数（IMF）分量占有较大的方差贡献,而低频分量相对较小;对各IMF分量的预测结果表明,起始阶段的高频IMF分量具有较好的可预测性,距平相关系数（fACC）值相对较高,高频分量的预测效果随预测时段加长而逐渐下降,均方根误差逐渐加大,低频分量的ACC值在起始阶段相对较低,随预测时段加长而逐渐加大,整个预测时段可预测性较强;最终的合成预测流量曲线表明,基于上述的思想,算法在20h时段的流量预测效果较好,拥有较高的ACC值和相对较低的均方根误差,30h时段的同号率均较为理想.

英文摘要：

This paper creates a time-based air traffic flow prediction model through the comprehensive application of ensemble empirical mode decomposition （EEMD） and least squares support vector machine（LS-SVM）. According to the EEMD results, the high frequence IMF components make greater variance contribution compared with the low frequency components. And the prediction results of different IMF components show that the high frequency IMF components are more predictive at the initial stage with a relatively high anomaly（ACC）,and that as the prediction period gets longer,the prediction effect of the high frequency components decreases with the root-mean-square error increasing gradually. And the low frequency components demonstrate a relatively low ACC at the initial stage and the longer the prediction is, the bigger the ACC is, and it shows relatively strong predictive ability throughout the prediction period. At last, the combined flow prediction curve demonstrates that the algorithm based on the above methods works well in predicting traffic flows with a relatively high ACC and low RMSE within 20 hours and maintains satisfactory ratios of the same symbol within 30 hours.