Application Research On Monte Carlo Simulation And Optimization Of Decision Analysis Under Risk

"Management is decision." An entire decision process consists of three phases, which are intelligence activity, design activity and choice activity. Decision Making plays a key function in the process. The universal uncertainty is the root of risk. Decision analysis under risk has been an active research field of decision theory.While discarding faultful gradient methods traditionally employed in decision analysis under risk, this thesis combines modern meta-heuristic algorithms like Genetic Algorithm and Tabu Search with Monte Carlo simulation to construct hybrid optimization algorithms. Simulation and optimization are implemented on Microsoft Excel platform to solve the stochastic model.After analyzing uncertainty and risk, a decision model under risk is built using stochastic variables figured as probability distributions. As for the simulation outcome, various statistics, overlay chart, trend chart, tornado chart and spider chart of the objective forecasts are introduced. Correlation coefficients and variance contributions are ranked to produce sensitivity charts.The hybrid strategies use self-adaptive operators and Latin Hypercube sampling techniques to enhance the parallel capability and efficiency. Constrained optimization issue is also covered and four penalty functions are presented in this thesis.Application research and Benchmark problems evaluation reveal that Monte Carlo simulation and optimization on Excel platform provide a flexible and efficient modeling instrument for decision analysis under risk. The hybrid strategies exert the powers of both simulation and meta-heuristic optimization. Compared with the hybrid of TS and simulation, the hybrid of GA and simulation converges tardier but more beseems large-scale optimization problems.