Article ID: | iaor20041546 |
Country: | China |
Volume: | 34 |
Issue: | 4 |
Start Page Number: | 455 |
End Page Number: | 458 |
Publication Date: | Jan 2002 |
Journal: | Journal of Harbin Institute of Technology |
Authors: | Sun Baiqing, Wang Xiaowei, Feng Yingjun, Wang Xuefeng, Pan Qishu |
Keywords: | neural networks |
The traditonal neural network has slow convergence speed and may easily fall into local minimum points, and its classification rule is not consistent with its training rule. All these influence the performance of neural network in pattern recognition. A region mapping model is proposed for decision on termination of an R and D project, which guarantees inherently the consistency of classification rule with the training rule. The convergence speed of region mapping model is faster than that of the neural network, and the model has a good identification ratio. The result of an empirical study shows that region mapping model is an effective method of pattern identification for decision on termination of R&D project.