| 研究課題/領域番号 |
26730113
|
|---|
| 研究機関 | 東京大学 |
|---|
研究代表者 |
宋 軒 東京大学, 空間情報科学研究センター, 特任准教授 (20600737)
|
|---|
| 研究期間 (年度) |
2014-04-01 – 2016-03-31
|
| キーワード | Human Mobility / Disaster Informatics / Data Mining / Big Data Application |
|---|
| 研究実績の概要 |
In 2014 facial year, our achievements of this project can be summarized as follows: (1) An intelligent system for urban emergency management during large-scale disasters was developed that automatically learns a probabilistic model to better understand and simulate human mobility during emergency situations. Based on the learned model, population mobility in various urban areas impacted by the earthquake throughout Japan can be automatically simulated or predicted. (2) An HMM-based human behavior model and urban mobility model was developed to accurately predict human behavior and mobility following large scale disasters. (3) A spectrum based approach was developed to effectively understand human disaster activities in a city. (4) A novel knowledge transfer model was developed for simulating a large amounts of human emergency behavior and mobility in any kind of disaster conditions. These research results were published in the eminent publications for computer science including KDD 2014, UbiComp 2014, AAAI 2014 and 2015.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
1: 当初の計画以上に進展している
理由
In generally, the expected goals of this project have been accomplished more than 70%, and the research progress is very good. In this facial year, we have established several powerful models or approaches to model, understand and predict human emergency behavior and mobility following natural disasters, such as HMM-based models, spectrum approach, etc. The various kinds of experiments and evaluations shows the superior performance of them. We also published many high quality publications for computer science including KDD 2014, UbiComp 2014, AAAI 2014 and 2015.
|
| 今後の研究の推進方策 |
In next fiscal year, our research activities will focus on the following aspects: (1) Because our collected data was big and heterogeneous, we found that with the increasing amount of training data, the performance of our model will face some bottlenecks. In the new facial year, we will try to build up Deep Belief Net and utilize the deep learning technology to model large amount of human emergency movements. (2) In the new facial year, we will construct an intelligent system that is able to automatically predict and simulate human emergency behavior and mobility for any place and disaster condition.
|
| 次年度使用額が生じた理由 |
In this year, we focus on models development and algorithm evaluation, and the progress was much better than expected. Thus, we did not purchase additional equipment and recruit research assistance. Meanwhile, we publish many papers in many top conferences of computer science, so we used more money for the travel.
|
| 次年度使用額の使用計画 |
In next fiscal year, the additional budget will cover the following issues: (1) Equipment purchase: We will purchase some computer severs to deal with the computation problems. (2) Publication and Academic Conferences fee: In the new fiscal year, we will publish several papers in top journal or conferences to report our results. Hence, some parts of budgets will be needed to cover these fee. (3) Experiment Labor Fee: In next fiscal year, we need to carefully test and evaluate our system. Hence, some additional experiments and labor fee will be needed.
|
