An Online Adaptive Boosting Ensemble Approach to Human Mobility Prediction at a Metropolitan Scale
| Project/Area Number |
20K19782
|
|---|
| Research Category |
Grant-in-Aid for Early-Career Scientists
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
Basic Section 60060:Information network-related
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
FAN ZIPEI 東京大学, 空間情報科学研究センター, 特任講師 (70835397)
|
|---|
| Project Period (FY) |
2020-04-01 – 2023-03-31
|
| Project Status |
Completed (Fiscal Year 2022)
|
| Budget Amount *help |
¥4,030,000 (Direct Cost: ¥3,100,000、Indirect Cost: ¥930,000)
Fiscal Year 2021: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
Fiscal Year 2020: ¥2,730,000 (Direct Cost: ¥2,100,000、Indirect Cost: ¥630,000)
|
|---|
| Keywords | Emergency management / Spatial data mining / IoT in Urban System / Mobility Prediction / Ensemble Model / Human Mobility Modeling / Ubiquitous Computing / Urban Computing / Deep Learning / Transfer Learning |
|---|
| Outline of Research at the Start |
Irregular human mobility is critical to a smart city. In this proposal, we model the irregularities in three directions: 1)precedent irregular mobility is modeled as few-shot learning to build human mobility deep learning model from very few samples; 2)To predict an online unprecedented human mobility;, irregular mobility at individual-level is discovered and aggregated at a metropolitan scale via personalized human mobility modeling; 3)online model fusion (regular mobility predictor, precedent/unprecedented human mobility predictors) is conducted in an adaptive ensemble learning framework.
|
| Outline of Final Research Achievements |
I finished the development of the online adaptive ensemble mobility prediction model in the proposal and published several papers in either the proceedings of the top conference or transactions of the top journals related to the key methods in this project. For irregular traffic prediction, I proposed a curriculum learning method that separates the irregular traffic nodes from regular ones to enhance the predictability of irregularities and published in AAAI 2023. For the irregularity simulation part, I have published two journal papers, the top visualization journal IEEE TVCG and one ACM TSAS. For the online ensemble part, one journal paper has been accepted to the top data mining journal IEEE TKDE. In addition, one paper on the real time implementation on the mobility prediction system and some simulation results on the irregular events such as stoppage of the Musashino railway line has been accepted to the top spatial data conference ACM sigspatial.
|
| Academic Significance and Societal Importance of the Research Achievements |
規則的な移動パターンは、モデル化しやすく、予測しやすい。しかし、花火大会やコミケなどの集客イベントや、大地震や台風などの自然災害時の移動など、現実世界の多くの移動パターンには不規則性が見られます。防災や災害対策など、多くの応用シーンにおいて、人間の不規則な移動は、日常的な移動よりもはるかに重要な役割を担っています。このような背景から、本プロジェクトでは、不規則な移動のモデリングに焦点を当て、不規則な移動の特徴を様々な側面から分析し、適応的なアンサンブルアプローチによって統合します。
|
Report
(4 results)
Research Products
(20 results)
