Evolutionary knowledge discovery to create explainable knowledge from large data sets

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K11964
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 61030:Intelligent informatics-related
• Research Institution: Gunma University (2021-2022); Fukuoka Nursing College (2020)
• Principal Investigator: Shimada Kaoru 群馬大学, 情報学部, 教授 (20454100)
• Co-Investigator(Kenkyū-buntansha): 荒平 高章 九州情報大学, 経営情報学部, 准教授 (30706958)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: 知識発見 / データマイニング / 知識創出 / 進化計算 / アイテム集合 / 説明可能性

Outline of Final Research Achievements

In this study, we proposed a local knowledge discovery method to discover knowledge from large-scale data that is thought to best explain the characteristics of individual cases without going through the process of conventional model building. We defined an itemset with statistical characteristics as an explanatory local knowledge representation, and evaluated their properties, reliability, and reproducibility in implementing the discovery. We also developed a method for global knowledge creation from the local knowledge set that was discovered, and developed clustering techniques that are different from conventional methods, and techniques that attempt to express global knowledge by combining small groups of statistical characteristics. It was found to have the potential for knowledge creation from explanatory and individual characteristics using publicly available medical data and other data.

Free Research Field

知能情報学

Academic Significance and Societal Importance of the Research Achievements

説明可能な局所的知識表現として提案した統計的な特徴を背景として持つアイテム集合（ItemSB：Itemsets with Statistically Distinctive Backgrounds）はデータ分析におけるアイテム集合ベース手法と統計学的方法を橋渡しして扱うことができるため、大規模データの分析に統計学的手法を効果的に導入可能とする技術と位置付けられる。大規模データの分析を統計的な背景をもつ小集団の組合せで扱おうとする新たなクラスタリング手法や、小集団にみられる統計的な特徴の連結により大域的な知識を扱おうとする方法の開発などの実用的で発展性のある独自方式による技術開発である。