Methods for selecting and testing hypothesis in big data-driven science and its demonstration in materials, biology, and medicine
| Project/Area Number |
17H00758
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (A)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Nagoya Institute of Technology |
|---|
Principal Investigator |
Takeuchi Ichiro 名古屋工業大学, 工学(系)研究科(研究院), 教授 (40335146)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
二宮 嘉行 統計数理研究所, 数理・推論研究系, 教授 (50343330)
豊浦 和明 京都大学, 工学研究科, 准教授 (60590172)
安河内 彦輝 三重大学, 地域イノベーション推進機構, 助教 (60624525)
井上 圭一 東京大学, 物性研究所, 准教授 (90467001)
|
|---|
| Project Period (FY) |
2017-04-01 – 2021-03-31
|
| Project Status |
Completed (Fiscal Year 2020)
|
| Budget Amount *help |
¥42,510,000 (Direct Cost: ¥32,700,000、Indirect Cost: ¥9,810,000)
Fiscal Year 2019: ¥12,350,000 (Direct Cost: ¥9,500,000、Indirect Cost: ¥2,850,000)
Fiscal Year 2018: ¥12,350,000 (Direct Cost: ¥9,500,000、Indirect Cost: ¥2,850,000)
Fiscal Year 2017: ¥11,700,000 (Direct Cost: ¥9,000,000、Indirect Cost: ¥2,700,000)
|
|---|
| Keywords | 機械学習 / 統計科学 / 材料科学 / 生物化学 / 医療科学 / Post-Selection Inference / 生物科学 / Selective Inference / ビッグ・データ / ビッグデータ |
|---|
| Outline of Final Research Achievements |
In various fields of scientific research, it has become possible to measure vast amounts of data about the research subject. The approach to scientific discovery based on such data is called data-driven science. In data-driven science, hypotheses are selected based on the data, but there is a risk that hypotheses that are over-fitted to the data may be selected incorrectly, and reliability evaluation of the data-driven hypotheses must be conducted appropriately. In this study, we established and demonstrated a method for evaluating the reliability of data-driven hypotheses in the fields of materials, biology, and medicine using a technique called selective inference.
|
| Academic Significance and Societal Importance of the Research Achievements |
研究対象から得られるデータに基づいて科学的発見を目指すアプローチはデータ駆動型科学と呼ばれ，さまざまな分野で有望視されている．しかしながら，データから仮説を選択する際に選択バイアスが生じてしまい，特に，誤った意思決定が重大なリスクとなる分野においては，データ駆動型仮説の信頼性評価が不可欠である．本研究ではデータ駆動型仮説の信頼性評価を行うための方法論を確立し，これをさまざまな分野で実証した．本研究の成果は健在なデータ駆動型科学の発展に寄与するものである．
|
Report
(4 results)
Research Products
(23 results)
