Abstract: This paper uses the link prediction method to analyze the route network of small and medium-sized airports and establishes four prediction proximity indexes considering network exogenous attributes and endogenous factors. The prediction effects of each index are compared and three coupling proximity prediction algorithms are designed. The algorithm with the highest prediction accuracy is selected to predict the new routes of small and medium- sized airports for the future condition. The prediction results are then verified. The results show that the prediction accuracy of the four endogenous factor indexes is higher than that of the exogenous attribute indexes except takeoff and landing sorties, among which the prediction accuracy of the local path (LP) index is the highest. Mining the internal structure of the network is more effective to predict the network connection, and the prediction effect of the coupling algorithm is better than that of a single index. Comparing the prediction results based on the coupling algorithm with the actual planed new routes in the future, the coupling algorithm actually predicted more than one third of the new routes in the future for small and medium-sized airport routes. The predicted routes are mainly concentrated in the eastern region, and most small and medium-sized airports will still choose to connect with central cities. The results are consistent with the small and medium-sized airport connections in real situation.

Key words: air transportation, proximity index, link prediction, small and medium-sized airports, airline network

摘要： 以中小机场航线网络为研究对象，利用链路预测理论与评价方法，从网络外生属性与内生因素角度各自建立了4种预测接近性指标，比较各指标的预测效果，并进一步设计出3种耦合接近性预测算法，选取预测精度最好的算法预测中小机场未来新增航线，并对预测结果进行实际检验。结果显示，4种内生因素指标的预测精确度均高于除起降架次以外的外生属性指标，其中局部路径(Local Path，LP)指标的预测精确度最高，挖掘网络内部的结构信息更有利于预测网络连接情况，耦合算法的预测效果优于单个指标。基于耦合算法的预测结果与未来新增航线对比发现， 预测的中小机场航线在未来新增航线中占比高于1/3，预测航线主要集中在东部地区，多数中小 机场依然会选择中心城市连接，结果较为符合中小机场实际连接情况。

关键词: 航空运输, 接近性指标, 链路预测, 中小机场, 航线网络