Analyzing And Modeling Express Shipping Service Network Based On Logistics Data

Along with the increasing prosperity of market economy and the growth of online retail, ex-press shipping service (e.g. EMS, SF Express) is playing an increasingly important role in our daily lives. As the complex network theory and big data analytics is becoming more developed, it is possible to understand express shipping service network (ExpressNet), analyze ExpressNet and solve the problems that express industries are facing through logistics big data. On the one hand, a thorough studying on the ExpressNet topology and the package traffic dynamics is essential for per-formance evaluation, network optimization and income growth on the express company’s side. On the other hand, precisely estimating the package delivery delay is not only helpful for the express companies to improve their services, but also brings great convenience for the customers. In this thesis, an universal research architecture and systematic work are proposed to study ExpressNet and thus characterize and model its package traffic dynamics based on logistics trace data.The platform for logistics data analysis is first introduced in this work, and the further analysis and modeling of ExpressNet are all carried out based on this platform, which includes logistics data crawling, data processing through Hadoop distributed computing system, and etc. We collect 16 million delivery traces and implement Hadoop, an open source distributed computing architecture, to process the data set. Furthermore, we reveal the topology structure of ExpressNet, analyze the traffic spatial-temporal dynamics, and show the package delay distributions. We find the skewed distribution of node degree as well as path length. The traffic flow are also uneven distributed and only a small part of nodes and paths bear the majority of the traffic flow. The package traffic show strong diurnal and weekly characteristics, which reveals people’s daily and weekly working schedule. Due to complicated package traffic dynamics, we propose an extended Markov model (EMM) to depict the package delivery process and further predict the package delay, and the data-based evaluation shows that EMM has a high prediction accuracy. At last, the procedure of using EMM to predict the package delay is given and the strategy to periodically update the model parameters with new logistics data is brought out to ensure the high performance of EMM.The results obtained by this work can not only benefit the express companies to better under-stand the ExpressNet and provide the theoretical support to improve the network performance, but can also help the customers fully enjoy the express services. What’s more, the data-driven study in this thesis can bring some insights for other similar data-driven work.