计算机技术与发展2017，Vol.27Issue(4)：1-5,5.DOI:10.3969/j.issn.1673-629X.2017.04.001
基于DBN的金融时序数据建模与决策
Modeling and Decision-making of Financial Time Series Data with DBN
摘要
Abstract
In analysis of the financial time series data,some complex nonlinear systems are often encountered.It is difficult to accurately model the state equation of these complex systems with mathematical methods.Faced with the current problem of complexity and uncertainty of financial time series analysis,simulations of complex nonlinear systems has been translated into pattern recognition of financial time series data and various patterns of financial time series curves,such as ascending,declining and random,have been determined.By taking use of the advantages of deep learning in unstructured data processing,an improved financial time series modeling and analysis method with improved Deep Belief Network (DBN) decision-making algorithm has been proposed,by which time series data have been transformed into unstructured data to be taken as input of input layer training model for in-depth learning network and to use trained model to predict the financial transaction data sample selection.Experimental results show that the accuracy rate acquired by improved deep belief network method has been achieved by 90.544 2 percent in quantitative analysis of final samples.关键词
深度信念网络/受限玻尔兹曼机/深度学习/金融时序数据/预测与决策Key words
deep belief network/Restricted Boltzmann Machine (RBM)/deep learning/financial time series data/forecasting and decision分类
信息技术与安全科学引用本文复制引用
曾志平,萧海东,张新鹏..基于DBN的金融时序数据建模与决策[J].计算机技术与发展,2017,27(4):1-5,5.基金项目
国家自然科学基金资助项目(61271349) （61271349）
中国科学院科技服务网络计划(STS计划)项目(KFJ-EW-STS-140) （STS计划）
中科院先导项目(XDA06010800) （XDA06010800）
上海市科学技术委员会资助课题(14DZ1119100) （14DZ1119100）
