论文部分内容阅读
为了准确预测空中交通短期流量,减轻空管协调压力,基于K近邻算法构建了空中交通短期预测模型。首先,通过多次取K值比较相对误差来确定合适的K值。之后,对原有的K近邻模型进行改进,引入空间参数,提出了3种状态向量组合的K近邻模型:时间维度模型、向台航路-时间维度模型与时空参数模型。以某扇区雷达数据对该模型进行检测,结果表明:同时引入时空参数的K近邻模型误差最小,平均为14.16%;基于指数权重的距离衡量方式均能达到预测精度优化的效果;高斯权重预测法在时间维度模型下优于反函数法,引入空间参数则反之;指数权重距离下的反函数法预测的时空参数模型误差为13.94%。改进后的K近邻模型对不同流量情况都具有普适性,预测结果可为空中交通流量管理提供理论参考。
In order to accurately predict the short-term air traffic flow and reduce the air traffic control coordination pressure, a short-term air traffic forecasting model based on K-nearest neighbor algorithm is constructed. First of all, by comparing the relative error K value to determine the appropriate K value. Afterwards, the original K-nearest neighbor model is improved and the spatial parameters are introduced. Three K-nearest neighbor models of state vectors are proposed: time dimension model, heading route-time dimension model and spatio-temporal parameter model. The results show that the error of K nearest neighbor model with the spatiotemporal parameters introduced is the smallest, with an average of 14.16%. The distance measurement based on exponential weight can all achieve the prediction accuracy optimization effect. The Gaussian weight prediction The method is better than the inverse method in the time dimension model, but the spatial parameter is the opposite. The error of the spatiotemporal parameter model predicted by inverse function under the exponential weight distance is 13.94%. The improved K-nearest neighbor model is universal for different flow conditions, and the prediction results can provide theoretical reference for the management of air traffic flow.