The Research On Risk Management Of Project Based On Aritificial Neural Network

Risk always appears during the course of project management and greatly affects the cost, time and quality of the project. So it's very important to carry on active management to the risk, to exactly identify and assess the risk in time, to dispose risk effectively and to decrease the risk loss to the lowest level. But at the same time some risk questions can't be solved using traditional statistical method and mathematical method. We must search a new way to ascertain and express the nonlinear relation between risk factors and risk loss. This paper proposes the thought to carry on risk management with ANN method after analyzing the primate content of project risk management and the characteristics of ANN.Because project risk management is a large system, this paper only study the use of ANN method in two aspects, integrated risk assessment and risk management of the engineering changes and claims. The author sets up a BP network model to assess integrated risk and a RBF network model to forecast engineering change and claim rate. Finally, the author builds the framework of ANN-based expert system of project risk management and the system's structure taking the integrated risk assessment for example.The paper draws the conclusion as that:(1) We can identify risk, forecast risk, assess risk and supply evidence to risk decision by the use of ANN.The BP network model to assess integrated risk is accurate and can be used to assess integrated risk of the projects which are similar to the training samples. But how to select the assessment indexes and change them into numerical information should be discussed. The RBF network model can be applied, but the range of application is limited because the risk factors change with the change of risk undertaker, project type and the project site.(2) The development of risk management expert system based on ANN is significative and feasible. The knowledge base of the ANN -based risk management expert system on contains neural network's...