摘要： 利用BP神经网络的并行性、容错性、自学习的特点,尤其是以任意精度逼近任何连续的非线性函数的功能,针对箱体振荡股票进行短期股价预测,同时引入了另一个诊断神经网络,评估预测神经网络的预测可信度,综合拟定操作策略。仿真试验表明,该方案是可行的。

关键词: 价值预测, 神经网络, 仿真, 自学习

Abstract: This paper makes use of the excellent characteristics of BP neural networks such as: parallel processing, fault tolerance, weight-learning, and especially nonlinear-functional approximation, to realize short-term prediction of a kind of stock price which fluctuates in a box, using the modified error back propagation(BP) and a weight-learning algorithm . At the same time, another diagnostic neural network is introduced to diagnose the credibility of the first neural prediction and makes decisions. The simulation results show the system is effective.

中图分类号: 

• TP183