遗传BP神经网络在大宗商品市场风险预警中的应用

Application of GA-BP Neural Network in the Risk Early-Warning of Bulk Commodity Electronic Trading Marketplace

研究利用遗传BP神经网络预警大宗商品电子交易市场风险的应用方法,将定量分析的思维方式引入大宗商品市场风险评价管理中.为此目的,建构了一个基于遗传BP神经网络的预警模型(GA-BPNNM),在市场调研的基础上建立了大宗商品电子交易市场风险评价指标体系,并通过实验确定了预警模型的最佳训练函数和隐层的最佳节点数.GA-BPNNM借助BP神经网络强大的自学习能力和非线性映射能力,克服传统手段在分析大宗商品电子交易市场风险时因其定义的模糊性和诱发因素的多样性所带来的困难;同时通过遗传算法与BP网络两者相互融合优化,解决BP神经网络易落入局部最优、收敛速度慢以及遗传算法易早熟等问题.仿真测试实验表明,GA-BPNNM预测结果优于标准BP神经网络预测方法,用于大宗商品电子交易市场风险损失程度预警是有效可行的.

The application of genetic BP neural networks used in early warning of the risk of bulk commodity electronic trading marketplace is studied. According to the risk characteristics of bulk commodity electronic trading marketplace, an early warning model（GA-BPNNM） based on genetic algorithm and BP neural networks is being built by the Matlab toolbox functions. The risk assessment index system is established based on the marketing research and the optimal training functions and number of nodes of the hidden layers are determined by the experiment. With the combinatorial optimization of genetic algorithm and BP neural networks, the difficulty of traditional risk analysis because of the ambiguity of definition and variety of causes could be overcome and the problems of premature phenomenon of genetic algorithm falling into local minima and slow convergence speed of BP neural network could be solved. The results of simulation test show that the GA-BPNNM prediction is better than BP neural networks and is feasible and effective in the early risk warning of bulk commodity electronic trading marketplace.