(A) study for development of interactive decision tree system

Abstract

Decision-Tree analysis is a convenient way to solve the alternative selection problem with uncertainty. Main purpose of this research is to study basic system designs for Decision-Tree and construct Decision-Tree System itself. Decision-Tree System design is composed of four procedures, i.e., input build, evaluation, output, alter procedures. In input build procedure, input data are stored in the VSAM-KSDS(Virtual Storage Access Method - Key Sequenced Data Set) file for the purpose of efficient autosorting and direct key retrieval. Especially in evaluation procedure, the meaningful processes for decision analysis such as EVPI(Expected Value of Perfect Information) and Sensitivity Analysis are represented. Top-Down approach is adopted to overall system design and PL/I is used as a programming language. For interactive system conversion, CLIST under TSO is used and detail system flow charts are also presented.