研究課題/領域番号 |
21K17746
|
研究種目 |
若手研究
|
配分区分 | 基金 |
審査区分 |
小区分60080:データベース関連
|
研究機関 | 東京工業大学 |
研究代表者 |
Le Hieu・Hanh 東京工業大学, 情報理工学院, 助教 (60813996)
|
研究期間 (年度) |
2021-04-01 – 2024-03-31
|
研究課題ステータス |
交付 (2022年度)
|
配分額 *注記 |
4,420千円 (直接経費: 3,400千円、間接経費: 1,020千円)
2023年度: 1,170千円 (直接経費: 900千円、間接経費: 270千円)
2022年度: 1,170千円 (直接経費: 900千円、間接経費: 270千円)
2021年度: 2,080千円 (直接経費: 1,600千円、間接経費: 480千円)
|
キーワード | data mining / privacy / medical data / differential privacy / recommendation |
研究開始時の研究の概要 |
This study aims to present a method for eliminating the need for trust in SPM while preserving privacy and providing secure, precise, and fast sequential data analysis that carefully learns the data distribution. The execution time should be reduced via parallel computation that utilizes modern hardware such as scalable multi-core CPUs. The feasibility of the proposed method will be studied using both open datasets and real medical data.
|
研究実績の概要 |
This study aims to present a method for eliminating the need for trust in sequential pattern mining (SPM) while preserving privacy and providing secure, precise, and fast sequential data analysis which carefully learns the data distribution. The fundamental algorithms of sequential data analysis on sequential medical data without privacy-preserving have been studied this year. In detail, several methods to analyze the sequence variants from more than one hospital have been designed and evaluated. The basic privacy-preserving SPM has also been studied in detail and the initial experimental results have been observed. For estimating the frequency of the sequences, an appropriate amount of noise is added to the original frequency to ensure privacy.
|
現在までの達成度 (区分) |
現在までの達成度 (区分)
3: やや遅れている
理由
The basic algorithm has been developed. However, evaluation process has been delayed due to the constraint of using sensitive data.
|
今後の研究の推進方策 |
The evaluation process will be improved by using other datasets which are easier to use. Moreover, further improving the performance of the privacy-preserving data analysis will be studied.
|