Robustness is a critical characteristic of ubiquitous real networked systems which are often very vulnerable to the random failures or targeted attacks. In this work, we aim to evaluate the robustness of a traffic network which is constructed from real trajectory data. First, we divide a city into grids which are considered as nodes and a link is established if there is at least one taxi dispatch order from one grid to another. Second, we propose two categories of attack strategies: node attacks (e.g. random node failure, and degree, in-degree, out-degree, closeness, clustering coefficient, PageRank-based targeted attacks) and link attacks (e.g. random link failure and link weight, link degree, link betweenness-based targeted attacks). Third, we suppose all elements (i.e. node or link) are simultaneously or iteratively removed, and evaluate the network robustness by employing the relative size of the giant components, average clustering coefficient and average shortest path length. The results show that the degree-based targeted attacks are the most destructive among all attacks, and no matter node or link, the betweenness-based attack is the most effective method to be beneficial to the target index. Moreover, most methods with recalculations take priority over the ones without recalculations.

中文翻译：

---

复杂网络视角下真实交通网络的鲁棒性评价

鲁棒性是无处不在的真实网络系统的一个关键特征，这些系统通常非常容易受到随机故障或有针对性的攻击。在这项工作中，我们旨在评估由真实轨迹数据构建的交通网络的鲁棒性。首先，我们将一个城市划分为多个网格，这些网格被视为节点，如果从一个网格到另一个网格至少有一个出租车调度订单，则建立链接。其次，我们提出了两类攻击策略：节点攻击（例如随机节点故障，以及度、入度、出度、接近度、聚类系数、基于PageRank的针对性攻击）和链接攻击（例如随机链接故障和链接权重、链接度、基于链接介数的针对性攻击）。第三，我们假设所有元素（即节点或链接）同时或迭代删除，并通过使用巨型组件的相对大小、平均聚类系数和平均最短路径长度来评估网络的鲁棒性。结果表明，基于度数的针对性攻击是所有攻击中破坏性最强的，无论是节点还是链路，基于介数的攻击都是有利于目标指标的最有效方法。此外，大多数重新计算的方法优先于没有重新计算的方法。