2017 Fiscal Year Annual Research Report
DeepMob: Learning Deep Models from Big and Heterogeneous Data for Next-generation Urban Emergency Management
| Project/Area Number |
17H01784
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
宋 軒 東京大学, 空間情報科学研究センター, 准教授 (20600737)
|
|---|
| Project Period (FY) |
2017-04-01 – 2020-03-31
|
| Keywords | Disaster Informatics / Big Data and Data Mining / Artificial Intelligence / Urban Computing / Internet of Things |
|---|
| Outline of Annual Research Achievements |
In the 2017 fiscal year, the research progress of this project is very good, and our research achievements can be summarized as follows:
(1)Firstly, we successfully collected big and heterogeneous data sources for the entire research. To manage these data, we developed a Hadoop cluster that consists of 32 cores, 32 GB memory, and 16 TB storage, and is able to run 28 tasks simultaneously.(2)Secondly, we build an online system called DeepUrban-Momentum to conduct the next short-term mobility predictions by using (the limited steps of) currently observed human mobility data. We apply our system to a real emergency scenario and demonstrate that our system is applicable in the real world.
Our research results were published in the eminent publications for computer science including AAAI 2018.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
In the 2017 fiscal year, the research progress of this project is very good. Firstly, we successfully collected big and heterogeneous data sources for the entire research.Secondly, we build an online system called DeepUrban-Momentum to conduct the next short-term mobility predictions by using (the limited steps of) currently observed human mobility data. We apply our system to a real emergency scenario and demonstrate that our system is applicable in the real world.
Our research results were published in the eminent publications for computer science including AAAI 2018.
|
| Strategy for Future Research Activity |
In the 2018 fiscal year, we will focus on developing novel approaches and deep models to predict human disaster behavior, mobility and preference. Meanwhile, we will also focus on the evaluation of the developed algorithms.
We will (1) use standard performance metrics used in the research literature and carefully compare our algorithms with existing work when applicable; (2) explore the trade-offs between competing alternative parameter settings through the design of suitable experiments; and (3) work closely with domain experts to evaluate and validate the results.
|
