对神经网络在评券投资管理三个阶段 :变量选择、收益预测、证券组合方面的应用进行了分析 .在此基础上 ,着重对不允许卖空情况下预期收益固定、风险最小的证券最优组合的投资比例系数的求解提出了一种确定性模拟退火神经网络的解法 .最后 ,通过实例验证了模型的有效性 This paper analyzes the application of neural network in the three stages of the investment management of securities investment: variable selection, income forecast and portfolio portfolio. On the basis of this, it emphasizes on the optimization of the securities with fixed expected return and minimum risk without short sale A solution to the deterministic simulated annealing neural network is proposed by solving the combined investment ratio coefficient.Finally, the effectiveness of the model is verified by an example