Research on the Seismic Direct Loss Fast Assessment Based on an Improved Neural Network

• Published in: 2014 Seventh International Symposium on Computational Intelligence and Design Vol. 2; pp. 449 - 454
• Main Authors: Weng Xun, Xiang Dao, Chen Xi
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2014
• Subjects: BP-ANN; Convergence; Earthquakes; Economics; fast assessment; Genetic algorithms; Neural networks; the data processing; Training
• ISBN: 9781479970049; 1479970042
• DOI: 10.1109/ISCID.2014.59

In this paper, the fast assessment model of seismic direct losses based on the genetic algorithm and advanced artificial neural networks is proposed, which is used to assess Earthquake-caused direct economic loss. The optimization of the initial weight and threshold by the application of genetic algorithms of BP neural network can avoid it getting into local minimum and obtaining effective result. To reduce both training time and the possibility of oscillation effectively, the additive momentum and self-adaptive-learn-rate adjustment method are adopted further to improve traditional BP algorithm. The results of training and testing the algorithm with the history statistical data show that the improved model not only needs shorter training time, but also has higher degree of precision and generalization ability. So it is more suitable to promote the application of the model in the seismic direct economic loss assessment.