2018 Fiscal Year Final Research Report
Analysis and visualization of sequential data for discovery without learning
| Project/Area Number |
16K12428
|
|---|
| Research Category |
Grant-in-Aid for Challenging Exploratory Research
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Multimedia database
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
Ohsawa Yukio 東京大学, 大学院工学系研究科(工学部), 教授 (20273609)
|
|---|
| Research Collaborator |
Hayashi Teruaki
Kasuga Akira
Iwasa Daiji
Yanaka Hitomi
Wang Jianshi
Ikegami Kenshin
Masui Norisada
Liu Keyang
Qi Ji
Yang Rui
Zhao Xiaoyi
Zeng Yanyuan
Yi Ji
Zhang Quexuan
Kishi Yoshiki
Iwanaga Hiroo
Yoshitaka Shitnaro
Emoto Mamoru
Takemura Kota
Naraoka Makoto
|
|---|
| Project Period (FY) |
2016-04-01 – 2019-03-31
|
| Keywords | 学習なき発見 / 時系列分析 / 変化の説明 / 潜在時間スケール / 変数選択 / 変数クエスト / 社会的ニーズ調査 |
|---|
| Outline of Final Research Achievements |
After grasping the need for qualitatively explaining the cause of change, it was confirmed that there is a need for technology to explain the cause of change from time series data in which various time scales are mixed. In line with this need, we established Tangled String (TS), which visualizes the start and end points of an entanglement (pill) as change points, and its extension method and evaluation method. In addition, while obtaining a group of algorithms that detect change points and explain them without learning, we also got results that produce machine learning methods that also explain the uncertainty of time series. We also developed Variable Quest that proposes just enough variables to be used in the method. The developed technology was applied to SNS, POS, and earthquakes to obtain results. We also surveyed the social needs of the outcome.
|
| Free Research Field |
データ市場設計学
|
| Academic Significance and Societal Importance of the Research Achievements |
データ利活用におけるニーズと、データと、データ分析技術等のマッチングを行うIMDJワークショップから、本研究の考え方で変化の原因を説明することへのニーズが実業界で強いことが明らかとなった。しかし、実際の社会・自然における変化には多様なタイムスケールが混在するため、従来手法では変化原因の説明が難しかった。この問題を解決することは基礎学術における進展であり、フェーズ①ニーズに答える技術的アプローチの検討, フェーズ②Tangled String(TS)を起点とする時間窓最適化, TSにかわるアルゴリズム群開発, フェーズ③ 実社会への適用と期待感の調査を行う手順により社会的意義も確保し続けた。
|
