IIS: A Multi-source Data Driven Optimization Framework for Interconnected Express Delivery System Design and Inventory Rebalance


Hui Xiong (hxiong@rutgers.edu) Principle Investigator

Award Information

Award Number1814510
DurationAugust 15, 2018 - July 31, 2021
TitleIIS: A Multi-source Data Driven Optimization Framework for Interconnected Express Delivery System Design and Inventory Rebalance
NSF Program(s) IIS

Project Abstract

The inter-connected express delivery system is very needed for many emerging applications, such as public bike rental service, electric car sharing service, and fresh product delivery. The successful deployment of inter-connected express delivery systems can greatly improve transportation, energy saving, food supply, and urban sustainability. Compared with traditional delivery systems, the inter-connected express delivery system has the following unique characteristics: (1) each station covers a small service area; (2) all stations are internally connected because they can act as inventories or suppliers to each other. There are two fundamental research challenges for the development of the inter-connected express delivery system: how to decide the station locations for a given area and how to timely rebalance the inventories among stations. It is very important to address these fundamental challenges in order to make the inter-connected express delivery system more effective, efficient and sustainable. This project aims to develop a data driven solution for solving these challenges. This study will advance the field of inter-connected express delivery system, expand the curricular content of data mining and optimization, and train undergraduate and graduate students.

This project focuses on two basic research problems: station site selection and station inventory rebalancing optimization. To solve the first problem, this project collects and analyzes a variety of data from different sources, such as historical demand data and geographic data, and combines neural network-based prediction method and combinatorial optimization techniques. To solve the second problem, this project identifies two distinct cases of the inventory rebalancing problem: static rebalancing and dynamic rebalancing. The research objective of the static rebalancing is to minimize the overall travel distance. This project develops a clustering-based heuristic solution for solving the static rebalancing in order to make the solution scalable for practical use. The research objective of the dynamic rebalancing is to minimize the overall unsatisfied demand, which involves much more uncertainty than the static one. This project develops a hybrid approach that combines advanced data mining and stochastic optimization techniques.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.


The results of this project have shown the commonality and the difference in the development of urban function zone. In particular, we have shown by integrating data mining techniques and spatial-temporal representation learning models, if properly exploited, can greatly improve the accuracy of urban regional mobility prediction. Last but not least, the results also reveal that it is necessary to take the multi-source data in the analysis of city zone functionality. In addition, the results of this project have provided a solution for incremental learning with deep models, and we provided a new way for mobile user profiling which is able to preserve inherent user activity patterns.

This project has also resulted in eight publications. One paper has been published in the top journal in the field of data mining: IEEE Transactions on Data and Knowledge Engineering (TKDE). Also, seven papers have been published in top conferences in the field of data, such as KDD and SIGIR.

Finally, one PhD student, Junming Liu, has graduated and joined the Department of Information Systems at the City University of Hong Kong as an Assistant Professor. Another PhD student, Yanchi Liu, has graduated as a Data Scientist at NCE Lab, in New Jersey. They have worked on a number of research problems produced in this project.

The proposed research has contributed significantly and creatively to the limited base of knowledge in the area of embedding learning in urban region function. Our finding, tools and documents could greatly help understand the regional human mobility and city area functionality over the world. Specifically, this project has the following major contributions in the discipline of urban region function. (1) This project provides an embedding framework for analyzing regional functions based on human mobility patterns, which have been applied to real-world in different cities. The results were found to be very useful by domain experts. (2) This project provides a clear understanding of the difference and the commonality in the development of diversified function zones. (3) In this project, we have revealed that integration of data mining techniques and optimization methods based on multi source data, if properly used, can greatly improve the performances of the embedding models.

The combination of data mining methods and optimization techniques for solving bike sharing problems will shed light on integrating multiple disciplines for other emerging applications. The developed bike station site selection techniques could broaden the application of optimization to other domains. The developed bike station re-balancing methods will open a new research direction in the field of transportation science.

This project has provided financial support for 1 graduate student who is pursuing his PhD study in smart bike sharing systems. This graduate student has completed his PhD study and joined the City University of Hong Kong as an Assistant Professor. Also, this project has produced several research problems which have been used as the dissertation topics by several graduate students. In particular, the dissertation topic of Dr. Yanchi Liu is on POI recommendations. He has recently graduated and joined the NCE Lab as a Data Scientist in NJ.


  1. Yang Yang, Da-Wei Zhou, De-Chuan Zhan, Hui Xiong, Yuan Jiang. Adversarial Substructured Representation Learning for Mobile User Profiling. The 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2019).
  2. Pengyang Wang, Yanjie Fu, Xiaolin Li, Hui Xiong. Adversarial Substructured Representation Learning for Mobility User Profiling. The 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2019).
  3. Can Chen, Yijun Wang, Guoan Huang, Hui Xiong. Hierarchical Demand Forecasting for Factory Production of Perishable Goods. The 2019 IEEE International Conference on Big Data (Big Data 2019).
  4. Yang Yang, De-Chuan Zhan, Yi-Feng Wu, Zhi-Bin Liu, Hui Xiong, Yuan Jiang. Semi-Supervised Multi-Modal Clustering and Classification with Incomplete Modalities. IEEE Transactions on Knowledge and Data Engineering (TKDE 2019).
  5. Zixuan Yuan, Hao Liu, Yanchi Liu, Denghui Zhang, Fei Yi, Nengjun Zhu, Hui Xiong. Spatio-Temporal Dual Graph Attention Network for Query-POI Matching. The 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2020).
  6. Shengming Zhang, Hao Zhong, Zixuan Yuan, Hui Xiong. Scalable Heterogeneous Graph Neural Networks for Predicting High-potential Early-stage Startups. The 27 th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2021).
  7. Denghui Zhang, Zixuan Yuan, Yanchi Liu, Hao Liu, Zuohui Fu, Fuzhen Zhuang, Hui Xiong, Haifeng Chen. Domain-oriented Language Modeling with Adaptive Hybrid Masking and Optimal Transport Alignment. The 27 th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2021).
  8. Zixuan Yuan, Hao Liu, Junming Liu, Yanchi Liu, Yang Yang, Renjun Hu, Hui Xiong. Incremental Spatio-Temporal Graph Learning for Online Query-POI Matching. The 28 th International World Wide Web Conference (WWW 2021).

© 2020 Data Mining Group @ Rutgers Business School