摘要
为提高工时定额的计算效率和准确度,通过对工时定额表的特点分析,设计开发了基于神经网络的工时定额系统,并针对BP神经网络存在的易陷入局部最小值、不易收敛、全局搜索能力差等缺点,将遗传算法集成到系统之中。介绍了系统的主要功能和工作流程,详细论述了隐层结点数目的计算、样本的自动划分、遗传算法优化、脱离MATLAB环境、神经网络模型的保存等关键技术。实际应用证明系统具有计算精度高、可以独立于MATLAB运行、通用性强等特点。
To improve computation efficiency and accuracy of man-hour quota, a man-hour quota system based on neural network was ex- plored through the analysis of the characteristics of the man-hour quota table. For the BP neural network has disadvantages like getting into lo- cal minimum point easily,low convergence speed, weak global search ability and so on, genetic algorithm was integrated into the system. The primary functions and work flow of the system were ir^troduced, and the key technologies were explained, including calculation of hide layer node, automatic sample division, genetic algorithms optimization, breaking away from the environment of MATLAB, the storage of neural net- work models. Practical application shows that the system has high computational accuracy and universality, and could run without the support of MATLAB.
出处
《计算机应用与软件》
CSCD
2010年第8期205-208,共4页
Computer Applications and Software
