An efficient inductive learning method for object-oriented database using attribute entropy

摘要: The data-driven characteristic of the Version Space rule-learning method works efficiently in memory even if training set is enormous. However, concept hierarchy each attribute used to generalize/specialize hypothesis a specific/general (S/G) processed sequentially and instance by instance, which degrades its performance. As for ID3, decision tree generated from order attributes according their entropies reduce number some paths. Unlike Space, ID3 generates an extremely complex when Therefore, we propose called AGE (A_RCH+OG_L+ASE_, where ARCH="Automatic geneRation Concept Hierarchies", OGL="Optimal Generalization Level", ASE="Attribute Selection Entropy"), taking advantages learn rules object-oriented databases (OODBs) with least learning features entropy. By simulations, found performance our algorithm better than both ID3. Furthermore, AGE's time complexity space are linear instances.