基于粗糙集约简算法的加注系统风险预测模型

摘要

A risk prediction model of BP neural network is produced based on rough set theory. First, the decision of risk source proportion is transferred into the evaluation of attribution import by attribution reduction arithmetic. The attribution import can be calculated under the rough set theory, which made the decision of risk proportion more external and reasonable. Then, a model of import system risk prediction is produced by the self-generation function BP neural network, in which can enhance the efficiency by contracted risk source as input. Examples show that this model has fine expansibility and less spending.%基于粗糙集理论,提出了加注系统风险预测模型:首先,应用属性约简算法,将加注系统风险源权重的确定问题转化为粗糙集理论中属性重要性的评价问题,通过计算得到加注系统各风险源的权重,从而使加注系统风险源权重的确定更具客观性和合理性;其次,采用BP人工神经网络的自学习功能,建立一个加注系统风险预测模型,将相对约简的风险源作为系统输入,可较好地提高预测模型的效率.实例表明,该模型具有良好的扩展性和较低的运行开销.