Types of Expert Systems Interpretation Systems Prediction Systems - PowerPoint PPT Presentation

Types of Expert Systems • Interpretation Systems • Prediction Systems • Diagnosis Systems • Design Systems • Planning Systems • Monitoring Systems • Debugging Systems • Repair Systems • Instruction Systems • Control Systems

Analytic Operations INTERPRET IDENTIFY PREDICT CONTROL MONITOR DIAGNOSE

Synthetic Operations CONSTRUCT SPECIFY DESIGN ASSEMBLE CONFIGURE PLAN

Classification Problem Solving Heuristic Matching Solution Data refinement abstractions Solutions Data Three basic steps in heuristic classification: •Data Abstraction (definitional; qualitative; generalization) •Heuristic Match •Solution Refinement

Heuristic Classification in MYCIN Heuristic Matching Gram negative Compromised infection host Immunosuppressed E. Coli Infection Leukopenia Low WBC WBC < 2.5

Heuristic Classification in SOPHIE Heuristic Matching Qualitative Behavior at a port values of ports in a module Quantitative circuit Component facult behavior Local circuit measurements

Classification versus Construction • The distinguishing feature of heuristic classification is that the solution can be enumerated in advance, e.g. in diagnostic phase of MYCIN, the program selects from a fixed set of offending organisms. • However, in many tasks, solution are constructed rather than selected, e.g. MYCIN’s therapy program recommends some combination of antibiotics at various dosages. • Constructing a solution involves some model that contains knowledge concerning constraints: (1) Constraints upon configuration of solution components (2) Constraints upon inputs and outputs (3) Constraints upon interactions between (1) and (2)