The Application of Genetic Fuzzy Neural Network in Project Cost Estimate

Abstract

Applications of neural network were widely used in construct project cost estimate. Aim at handling weakness of poor convergence and insufficient forecast, an improved fuzzy neural network method based on SOFM (self-organizing feature map) and GA (genetic algorithm) was proposed to replace the fashionable T-S fuzzy neural network. The method illustrated how to apply SOFM and GA to improve the fault such as poor convergence and insufficient forecast. After optimizing of T-S fuzzy neural network model, construct project cost estimate model had been built up. Finally, the model was set up with the purpose of comparing generalization ability by 18 examples and 2 testing samples. Comparing the simulation, a positive result was found that genetic fuzzy neural network had a better performance in reducing the forecast error and iterating times than BP, BP optimized by GA, GA-BP, fuzzy neural work. Therefore, this model is fit for handling construct project cost estimate.