以广义回归神经网络预测共同基金报酬.pdfVIP

第9卷 第4期 长安大学学报(社会科学版) Vo1．9 NO．4 2007年 12月 Journal of Chang’an University(Social Science Edition) Dec．2007 【应用经济学研究】 以广义回归神经网络预测共同基金报酬 潘 文 超 (兰阳技术学院 信息管理系，台湾 台北 104) 摘 要：鉴于近年来许多相关文献成功地运用广义回归神经网络进行财经方面的预测，以及国内共 同基金净值之预测与报酬率评估。通过搜集国内基金资料，以灰关联分析法进行各基金投资绩效 分析，挑选投资绩效良好的共同基金作为投资标的；再以广义回归神经网络建立预测模型，与灰预 测模型、多元回归模型进行预测能力及报酬率的比较分析。5种预测绩效评价指标、5组数据交互 验证散布图及报酬率分析表明：广义回归神经网络在预测能力及预测报酬率上均有很好的表现。 关键词：灰关联分析；灰预测；广义回归神经网络；多元回归模型；遗传算法 中图分类号：F7830．91 文献标志码：A 文章编号：1671-6248(2007)04—0055-04 Forecast for mutual fund returns with gerenal regression neural network PAN Wen—chao (Department of Information Management，Lanyang Institute of Technology，Taibei 1 04，Taiwan，China) Abstract：In recent years，there are many relevant documents that are successfully in general regression neural network for the financial sector forecasts．This paper adoptes the general regression neural network for the prediction of the net value of the domestic mutual fund and for evaluation of the return of the investment．The author picks out a lot of fund information at home，analysizes its investment performance with grey relational analysis，and select some good investment performances of mutual funds as investment targets． Through general regression neural network model，he sets up the prediction model and with grey prediction and multiple regression model，he conducts the comparative analysis on the accuracy of the prediction and the return rate．It is found that it is better tO predict the return rate with general regression neural network than with grey prediction and multiple regression mode1．On the basis of the evaluation of the 5 indexes of the perfor