CrossBag: A Bag of Tricks for Cross-City Mobility Prediction

CrossBag: A Bag of Tricks for Cross-City Mobility Prediction. Lee, J. & Chiang, Y. In Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Human Mobility Prediction Challenge, of HuMob'24, pages 55–59, New York, NY, USA, 2024. Association for Computing Machinery.

Paper doi abstract bibtex

Access to large-scale human trajectory data has significantly advanced the understanding of human mobility (HuMob) behavior for urban planning. However, these data are often concentrated in major cities, leaving smaller or less-monitored areas with limited information, undermining the performance of data-hungry machine learning models for HuMob prediction. This imbalance poses a challenge for cross-city mobility prediction, as many existing models are designed for single-city settings. To address this, we present CrossBag, a set of simple yet effective techniques to boost cross-city prediction. These techniques include context-aware spatiotemporal embeddings, masking types, and a progressive knowledge transfer method to incrementally adapt the target model while preserving useful patterns from the source model for stable cross-city transfer. Additionally, we propose a test-time trajectory refinement method using top-K guided beam search to prevent predictors from getting stuck in repetitive location predictions. We validate CrossBag on the large-scale multi-city dataset from the HuMob Challenge 2024, achieving a top-10 placement out of over 100 participating teams.

@inproceedings{10.1145/3681771.3699935,
  abstract = {Access to large-scale human trajectory data has significantly advanced the understanding of human mobility (HuMob) behavior for urban planning. However, these data are often concentrated in major cities, leaving smaller or less-monitored areas with limited information, undermining the performance of data-hungry machine learning models for HuMob prediction. This imbalance poses a challenge for cross-city mobility prediction, as many existing models are designed for single-city settings. To address this, we present CrossBag, a set of simple yet effective techniques to boost cross-city prediction. These techniques include context-aware spatiotemporal embeddings, masking types, and a progressive knowledge transfer method to incrementally adapt the target model while preserving useful patterns from the source model for stable cross-city transfer. Additionally, we propose a test-time trajectory refinement method using top-K guided beam search to prevent predictors from getting stuck in repetitive location predictions. We validate CrossBag on the large-scale multi-city dataset from the HuMob Challenge 2024, achieving a top-10 placement out of over 100 participating teams.},
  address = {New York, NY, USA},
  author = {Lee, JangHyeon and Chiang, Yao-Yi},
  booktitle = {Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Human Mobility Prediction Challenge},
  doi = {10.1145/3681771.3699935},
  isbn = {9798400711503},
  keywords = {Human mobility, Spatiotemporal, Transfer learning, Transformer},
  location = {Atlanta, GA, USA},
  numpages = {5},
  pages = {55–59},
  publisher = {Association for Computing Machinery},
  series = {HuMob'24},
  title = {CrossBag: A Bag of Tricks for Cross-City Mobility Prediction},
  url = {https://doi.org/10.1145/3681771.3699935},
  year = {2024}
}

Downloads: 0

{"_id":"ZzXQfajCXWBXK9Mhi","bibbaseid":"lee-chiang-crossbagabagoftricksforcrosscitymobilityprediction-2024","author_short":["Lee, J.","Chiang, Y."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","abstract":"Access to large-scale human trajectory data has significantly advanced the understanding of human mobility (HuMob) behavior for urban planning. However, these data are often concentrated in major cities, leaving smaller or less-monitored areas with limited information, undermining the performance of data-hungry machine learning models for HuMob prediction. This imbalance poses a challenge for cross-city mobility prediction, as many existing models are designed for single-city settings. To address this, we present CrossBag, a set of simple yet effective techniques to boost cross-city prediction. These techniques include context-aware spatiotemporal embeddings, masking types, and a progressive knowledge transfer method to incrementally adapt the target model while preserving useful patterns from the source model for stable cross-city transfer. Additionally, we propose a test-time trajectory refinement method using top-K guided beam search to prevent predictors from getting stuck in repetitive location predictions. We validate CrossBag on the large-scale multi-city dataset from the HuMob Challenge 2024, achieving a top-10 placement out of over 100 participating teams.","address":"New York, NY, USA","author":[{"propositions":[],"lastnames":["Lee"],"firstnames":["JangHyeon"],"suffixes":[]},{"propositions":[],"lastnames":["Chiang"],"firstnames":["Yao-Yi"],"suffixes":[]}],"booktitle":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Human Mobility Prediction Challenge","doi":"10.1145/3681771.3699935","isbn":"9798400711503","keywords":"Human mobility, Spatiotemporal, Transfer learning, Transformer","location":"Atlanta, GA, USA","numpages":"5","pages":"55–59","publisher":"Association for Computing Machinery","series":"HuMob'24","title":"CrossBag: A Bag of Tricks for Cross-City Mobility Prediction","url":"https://doi.org/10.1145/3681771.3699935","year":"2024","bibtex":"@inproceedings{10.1145/3681771.3699935,\n abstract = {Access to large-scale human trajectory data has significantly advanced the understanding of human mobility (HuMob) behavior for urban planning. However, these data are often concentrated in major cities, leaving smaller or less-monitored areas with limited information, undermining the performance of data-hungry machine learning models for HuMob prediction. This imbalance poses a challenge for cross-city mobility prediction, as many existing models are designed for single-city settings. To address this, we present CrossBag, a set of simple yet effective techniques to boost cross-city prediction. These techniques include context-aware spatiotemporal embeddings, masking types, and a progressive knowledge transfer method to incrementally adapt the target model while preserving useful patterns from the source model for stable cross-city transfer. Additionally, we propose a test-time trajectory refinement method using top-K guided beam search to prevent predictors from getting stuck in repetitive location predictions. We validate CrossBag on the large-scale multi-city dataset from the HuMob Challenge 2024, achieving a top-10 placement out of over 100 participating teams.},\n address = {New York, NY, USA},\n author = {Lee, JangHyeon and Chiang, Yao-Yi},\n booktitle = {Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Human Mobility Prediction Challenge},\n doi = {10.1145/3681771.3699935},\n isbn = {9798400711503},\n keywords = {Human mobility, Spatiotemporal, Transfer learning, Transformer},\n location = {Atlanta, GA, USA},\n numpages = {5},\n pages = {55–59},\n publisher = {Association for Computing Machinery},\n series = {HuMob'24},\n title = {CrossBag: A Bag of Tricks for Cross-City Mobility Prediction},\n url = {https://doi.org/10.1145/3681771.3699935},\n year = {2024}\n}\n\n","author_short":["Lee, J.","Chiang, Y."],"key":"10.1145/3681771.3699935","id":"10.1145/3681771.3699935","bibbaseid":"lee-chiang-crossbagabagoftricksforcrosscitymobilityprediction-2024","role":"author","urls":{"Paper":"https://doi.org/10.1145/3681771.3699935"},"keyword":["Human mobility","Spatiotemporal","Transfer learning","Transformer"],"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"http://knowledge-computing.github.io/publications.bib","dataSources":["3sPtWLmmdPRfH69LS"],"keywords":["human mobility","spatiotemporal","transfer learning","transformer"],"search_terms":["crossbag","bag","tricks","cross","city","mobility","prediction","lee","chiang"],"title":"CrossBag: A Bag of Tricks for Cross-City Mobility Prediction","year":2024}