A number of recent disasters have challenged the functionality of transport networks. The significance of road transport infrastructure to the functioning means that systems need to be able to operate under undesirable conditions, and quickly return to acceptable levels of service. The objective of the study is to analyze real-world networks speed fluctuation and evaluate the quantitative relationship between resilience and graph-based metrics, and link attributes using crowd-sourced data. We measure resilience in terms of the rate (vehicle speed) at which the road network recovers from a disruptive event and define five metrics to quantify network resilience. We analyze more than 500 links affected by disruptions in multiple cities with more than millions of crowd-sourced data. Using changes in link speed before, during, and after the disruption, the resilience metrics are applied to three case studies that are categorized as no-notice disruption, notice disruption, and disruption caused by continuous events. The results indicate that link graph-based metrics and attributes have a high impact on network resilience. However, the relevance of different metrics and attributes to the link resilience is different. Population density, predictability of disasters, and human factors have a significant impact on the reduction and recovery phases.