首页>
外国专利>
A SYSTEM AND METHOD FOR DATA MINING FROM RELATIONAL DATABASES USING A HYBRID NEURAL-SYMBOLIC SYSTEM
A SYSTEM AND METHOD FOR DATA MINING FROM RELATIONAL DATABASES USING A HYBRID NEURAL-SYMBOLIC SYSTEM
展开▼
机译:使用混合神经符号系统从关系数据库挖掘数据的系统和方法
展开▼
页面导航
摘要
著录项
相似文献
摘要
escribed are a system and method for learning classification rules, (represented by a semantic network), from a database, using a hybrid neural-symbolic system. The objective of the invention is todescribe one attribute/variables, also occuring in data, and consequently apply this description to unseen, (i.e. not in data), cases. An initial domain knowledge is built from a database, representedby a semantic network. Then, this initial semantic network is transformed into a neural network. Next, this neural network is refined using a specific connectionist learning technique and the cases inthe database. Consequently, the refined neural network is used in order to revise the initial semantic network. Finally, the revised semantic network can classify new unseen cases with some classification accuracy.
展开▼