This project aims at supporting research for all aspects of spatiotemporal data modeling with machine learning and addressing many scientific, mathematical, industrial, and engineering problems in:
- Urban science & smart cities
- Human mobility modeling
- Geospatial data analysis
- Intelligent & sustainable urban systems
- Optimization & decision making
- Data standardization & valorization
- Signal processing
- Network science
- Graph computing & learning
- Causal inference & learning
- Computational social science
This project presents some spatiotemporal methods with data-driven machine learning:
- Xinyu Chen, HanQin Cai, Fuqiang Liu, Jinhua Zhao (2024). Correlating time series with interpretable convolutional kernels. arXiv:2409.01362.
- Xinyu Chen, Zhanhong Cheng, HanQin Cai, Nicolas Saunier, Lijun Sun (2024). Laplacian convolutional representation for traffic time series imputation. IEEE Transactions on Knowledge and Data Engineering. 36 (11): 6490-6502. [Slides] [Data & Python code]
- Xinyu Chen, Chengyuan Zhang, Xiaoxu Chen, Nicolas Saunier, Lijun Sun (2024). Discovering dynamic patterns from spatiotemporal data with time-varying low-rank autoregression. IEEE Transactions on Knowledge and Data Engineering. 36 (2): 504-517. [Blog post] [Data & Python code]
There are several easy-to-follow posts that are created to explain the essential ideas of this project:
- Time Series Convolution. A convolutional kernel approach for reinforcing the modeling of time series trends and interpreting temporal patterns, allowing one to leverage Fourier transforms and learn sparse representations. (Under development)
To advance the development of spatiotemporal data modeling in the research community, this project handles various spatiotemporal data:
- Analyzing millions of taxi trips in the City of Chicago
- Constructing human mobility tensor on NYC rideshare trip data
- Utilizing international import and export trade data from WTO Stats
To contribute to the open science, this project provides a series of tutorials for beginners to understand machine learning and data science, including
- Tensor decomposition for machine learning (see the detailed page at MIT Sites):
- Foundation of tensor computations
- Foundation of optimization
- Spatiotemporal data visualization in Python:
- High-resolution sea surface temperature data
- Global water vapor patterns
- Germany energy consumption
- Station-level USA temperature data
- Service-level mobile traffic data
For those who are interested in broad areas within the scope, we would like to recommend a series of well-documented reading notes.