APPLICATION OF NEURAL NETWORKS FOR THE EXTRACTION AND CHARACTERIZATION OF KNOWLEDGE CONTAINED IN DATABASES

LRGE DATABASES EXIST FOR USE BY MILITARY COMMANDERS AND EXECUTIVE DECISION-MAKERS. CONVENTIONAL TECHNOLOGY DOES NOT EASILY ALLOW THE EXTRACTION OR DISTILLATION OF THE KNOWLEDGE CONTAINED WITHIN THESE DATABASES. AS A CONSEQUENCE, MOST OF THE TRUE POWER OF THIS INFORMATION IS INEFFICIENTLY UTILIZED. HNC HAS DEVELOPED THE ABILITY TO CHARACTERIZE, EXTRACT, AND EXPLOIT KNOWLEDGE CONTAINED WITHIN DATABASES USING NEURAL NETWORK TECHNIQUES. THIS CAPABILITY HAS BEEN DEMONSTRATED ON COMMERCIAL SECTOR DATABASES, AND IS CALLED "KNOWLEDGE EXTRACTION USING NEURAL NETWORKS" (KENN). KENN CAN DISCOVER UNKNOWN RELATIONSHIPS CONTAINED WITHIN DATABASES, AND CAN DEVELOP COMPLEX NON-LINEAR MODELS OF THESE RELATIONSHIPS WITH NO PROGRAMMING AND MINIMAL USER INTERACTION. KENN IS EXTREMELY POWERFUL AND MAJOR PORTIONS OF THE TECHNOLOGY EXIST. HOWEVER, THE MOST POWERFUL COMPONENT OF KENN, RELATIONSHIP DISCOVERY, HAS ONLY BEEN DEMONSTRATED ON A PROOF-OF-CONCEPT BASIS. THIS PROPOSAL DESCRIBES THE EXISTING KENN TECHNOLOGY AND OFFERS A PLAN FOR THE ENHANCEMENT OF THE RELATIONSHIP DISCOVERY COMPONENT OF KENN WITH SPECIFIC EMPHASIS TO ARMY EXECUTIVE INFORMATION SYTEM (EIS) PROCESSING REQUIREMENTS. COMPLETION OF THIS PROPOSED EFFORT WILL RESULT IN AN EXTREMELY POWERFUL, AUTOMATED EIS TOOL TO ASSIST DECISION MAKING PROCESSES.