| Title |
A Study on Predicting Cost Estimation of Apartment Building Using Neural Network's Architecture Optimized by Genetic Algorithms |
| Authors |
Kim Gwang-Hee ; Kang Kyung-In |
| Keywords |
공사비 예측 ; 신경망 ; 유전자 알고리즘 ; 신경망 구조 최적화공사비 예측 ; 신경망 ; 유전자 알고리즘 ; 신경망 구조 최적화 |
| Abstract |
The purpose of this study was to propose the method of improving generalization of neural networks by optimizing NN's architecture and parameters using genetic algorithms. The models of NNs and regression analysis applied were error back-propagation network and linear multi- regression analysis, respectively. This study applied the historical data of apartment buildings' direct costs to each model for verifying the validity of it. The results of this study were as follows: (1) the model that NN's architecture and parameters were optimized by GAs was superior to the regression model in cost estimation of apartment buildings. (2) the application of genetic algorithms may be a proper solution to the problem that user have because of the lack of adequate rules for determining the parameters of neural networks in utilizing neural networks. |