Performance Improvement in the Implementation of DBLEARN
Colin Carter, Howard J. Hamilton · 1994
We describe efficiency improvements made to the DBLEARN program for performing knowledge discovery in databases. The original DBLEARN prototype implementation suffered from relatively poor performance. Causes of this inefficiency are explored and implementation strategies are suggested to improve performance. They include better memory management, more efficient storage of attribute values, and better searching techniques for matching attribute values with more general concepts. Finally measurements of the performance improvement are provided for one specific computer system. 1 Introduction Knowledge discovery from databases is the automated extraction of useful and interesting information from large bodies of diverse data stored in databases. The primary objective of such discovery is to produce new nontrivial information that is understandable, accurate and useful [Frawley et al., 1992]. One of the primary challenges of such discovery is to devise a method which computers can implem...