Knowledge-based Decision Tree Generation

Period of Performance: 01/01/1987 - 12/31/1987

$565K

Phase 2 SBIR

Recipient Firm

Perceptics Corp.
PO BOX 22991 - PELLISSIPPI
Knoxville, TN 37933
Principal Investigator

Abstract

A SYSTEM IS PROPOSED FOR THE UTILIZATION OF EXPERT KNOWLEDGE TO GENERATE HIGHLY EFFICIENT SOLUTIONS TO DECISION-MAKING PROBLEMS. STARTING FROM EITHER A PRODUCTION RULE OR A DECISION TREE REPRESENTATION OF THE PROBLEM, AS SUPPLIED BY A SYSTEM DESIGNER, THE EXPERT DECISION TREE GENERATOR SYSTEM CAN PRODUCE AN EFFICIENT DECISION TREE REPRESENTATION. THE USER CAN THEN EXPLORE SOME OF THE POSSIBLE OPTIMAL SOLUTIONS TO THE PROBLEM, CHOOSING THE CONSIDERED BEST SUITED FOR THE PARTICULAR APPLICATION. THE SYSTEM IS ALSO CAPABLE OF GENERATING HIGH-LEVEL LANGUAGE CODE (C OR ADA) THAT IMPLEMENTS A DECISION TREE REPRESENTATION. THE PURPOSE OF THIS INVESTIGATION IS TO DEVELOP AND IMPLEMENT A SET OF ALGORITHMS THAT CAN BE USED FOR THE SYNTACTIC MANIPULATION OF DECISION TREES AND THE GENERATION OF ALL THE OPTIMAL FORMS OF A DECISION TREE. THESE ALGORITHMS ARE ALSO CAPABLE OF HANDLING TERMS IN DECISION THEORIES WITH BOTH DEPENDENCIES AND VARIABLES. THE RESULT IS A PROBLEM-SOLVING APPROACH WHERE CONCEPTUAL AND EXPERT KNOWLEDGE RATHER THAN HEURISTIC ARE THE DRIVING FORCE IN THE DESIGN AND IMPLEMENTATION OF EFFICIENT DECISION-MAKING PROCESSES.