| 神经网络集成股票市场预测研究 |
| |
| 引用本文： | 姜波,吴建生.神经网络集成股票市场预测研究[J].烟台职业学院学报,2006,12(3):59-65. |
| |
| 作者姓名： | 姜波 吴建生 |
| |
| 作者单位： | 1. 山东工商学院,数学学院,山东,烟台,264005
2. 广西柳州师范高等专科学校,数学与计算机科学系,广西,柳州,545004 |
| |
| 摘 要： | 利用遗传算法的全局搜索能力,同时进化设计神经网络的网络结构和初始连接权,获得一组合适网络结构和初始连接权,再进行新一轮BP神经网络训练,生成神经网络的集成个体,采用二次规划计算方法,计算各集成个体的最优非负权系数进行最优组合集成,生成神经网络的输出结论,以此建立股市预测模型。通过上证指数开盘价进行实例分析,计算结果表明该方法预测精度高、稳定性好,易于操作。
|
| 关 键 词： | 遗传算法 神经网络 二次规划 |
| 文章编号： | 1673-5382(2006)03-0059-07 |
| 修稿时间： | 2006年4月11日 |
Study on the Stock Market Prediction Model Using the Quadratic Programming Optimal Combination of Neural Ensemble Based on Genetic Algorithms |
| |
| JIANG Bo,WU Jian-sheng.Study on the Stock Market Prediction Model Using the Quadratic Programming Optimal Combination of Neural Ensemble Based on Genetic Algorithms[J].Journal of Yantai Vocational College,2006,12(3):59-65. |
| |
| Authors: | JIANG Bo WU Jian-sheng |
| |
| Abstract: | This paper presents the evolving neural network architecture and beginning connection weights based on genetic algorithms in order to produce better network architecture and beginning connection weights;train again the traditional back propagation by training samples.The ensemble strategy is carried out that the quadratic pro-gramming to calculate the best non-negative weights.The weighted coefficient of each ensemble individual is obtained.The forecast model of Stock Market is established with this method.The result shows this method is of high accuracy and stability,and it is a useful model for the actual operational forecasting. |
| |
| Keywords: | genetic algorithms neural network quadratic program |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
