An Expert System for Stochastic Control Problems: Automatic Report Generation

Jean Philippe Chancelier, Claude Gomez, Jean Pierre Quadrat, Agnès Sulem

Research output: Contribution to journalArticlepeer-review

Abstract

We present a prototype expert system for the treatment of stochastic control problems. The objective is to automate all the steps involved in the solution of such problems, using computer algebra, inference techniques and symbolic manipulations. The system, written in Macsyma, Lisp and Prolog, accepts input in natural language and in symbolic form; it carries out the basic analysis of the problem, makes theoretical analyses to study the existence and uniqueness of solutions and selects a method among: Dynamic Programming, Optimization in the class of local feedbacks, Monte Carlo or Stochastic Gradient method, Perturbation method. The system generates a Fortran code for the numerical solution of the problem and finally a report including a description of the method and graphs for the numerical results. In this paper, we emphasize the automatic generation of the report.

Original languageEnglish
Pages (from-to)65-82
Number of pages18
JournalComputer Science in Economics and Management
Volume2
Issue number1
DOIs
Publication statusPublished - 1 Jan 1989

Keywords

  • Stochastic control problems
  • automatic report generation
  • expert system

Fingerprint

Dive into the research topics of 'An Expert System for Stochastic Control Problems: Automatic Report Generation'. Together they form a unique fingerprint.

Cite this