基于GA-BP神经网络的巨项目投入评价的改进研究

考虑到巨项目巨大的资源投入量、投入产出的多样性、复杂性和社会性等特点, 根据决策"核心三原则"构建适宜的投入产出评价指标体系, 通过DEA法对巨项目各投入方案进行初评并筛选出有效的方案, 在此基础上运用遗传算法优化的BP神经网络(GA-BP), 结合DEA的初评结果对多个有效方案进行二次评价, 从而选择最优方案. 实证分析表明, 此改进的巨项目投入评价模型, 能较好地适应巨项目的特征, 既保证了投入评价的客观性, 又能够实现投入方案的完全排序.

Improvement of input evaluation for giant projects based on GA-BP neural network

Considering a giant project's huge amount of resources input and diversity, complexity and social effect of its input and output, the input-output evaluation index system was established according to "three core principles". Each input scheme of the giant project was first evaluated by DEA method and effective input schemes were selected. Based on GA-BP neural network and results from first evaluation, multiple effective schemes were evaluated again to choose the optimal one. The empirical analysis shows that this improved evaluation model is adapt to characteristics of giant projects. It can not only guarantee the objectivity of giant projects input evaluation, but also achieve the goal to sequencing all input schemes.