分层逐步求精的知识发现KDD模型结构及其应用

Multi-Level and Incremental Structure of Knowledge Discovery in Database and Its Application

知识发现（ＫｎｏｗｌｅｄｇｅＤｉｓｃｏｖｅｒｙｉｎＤａｔａｂａｓｅ）作为人工智能、机器学习、模式识别、统计学、数据库和知识库、数据可视化等众多学科相互交叉融合发展起来的新兴研究领域，已引起了学术界和工业界的广大关注．本文提出基于示例学习分层逐步求精的ＫＤＤ模型结构，在借鉴模式识别有关知识的基础上，有机地与领域知识相结合，应用于无约束手写数字识别ＫＤＤ系统，得到较满意的结果．

Knowledge discovery in database (KDD) is a new research field which is being developed combining subjects of artificial intelligence, machine learning, pattern recognition,statistics, database and knowledge base, digit visualization, etc. KDD has brought aboutgreat influence in academia and industry. In this paper, a multi-level and incremental structure of KDD is presented, which uses technology of instance-based learning and knowledgeof pattern recognition. Combined with the knowledge of an actual domain and applied to therecognition of unconstrained handwritten numerals, this structure gets satisfying results.