摘要
通过有效融合与汇率相关的互联网搜索信息和宏观经济信息,提出一个新的汇率预测方法.一方面,根据信息丰富的互联网大数据,将选取的百度指数关键词信息合成能反映投资者关注度的百度综合搜索指数,再利用核主成分(KPCA)方法对宏观经济变量的信息进行提取,合成宏观综合影响指数,最后构建基于多源信息融合的汇率预测模型;另一方面,分别采用BP、KELM和SVM模型进行预测.为减小预测误差,对神经网络连接权重和阈值使用灰狼优化算法(GWO)进行了优化.通过对美元兑人民币汇率进行实证发现,融合多源数据信息之后,使用GWO-BP预测模型能获得更好的预测性能.
A new exchange rate forecasting method is proposed by effectively fusing exchange rate-related internet search information and macroeconomic information.On the one hand,according to the information-rich internet big data,the selected about:blank index keywords are synthesized into a network comprehensive search index which reflects the attention of investors.Then effective information extraction is performed on macroeconomic variables by using a kernel principal component analysis(KPCA)method to synthesize a macro comprehensive impact index.Finally,an exchange rate prediction model based on multi-source information fusion is constructed.On the other hand,BP,KELM and SVM models are used for prediction.In order to reduce the prediction error,the grey wolf optimization algorithm(GWO)is used to optimize the neural network connection weights and threshold parameters.By comparing the prediction results,it is found that the GWO-BP prediction model can obtain better prediction performance after fusing multi-source data information.
作者
刘艺彬
孙景云
Liu Yibin;Sun Jingyun(Lanzhou University of Finance and Economics;Center for Quantitative Analysis of Gansu Economic Development)
出处
《哈尔滨师范大学自然科学学报》
CAS
2023年第3期28-37,共10页
Natural Science Journal of Harbin Normal University
基金
甘肃省科技计划项目(21JR1RA280)
2022年陇原青年创新创业人才项目
关键词
网络搜索信息
宏观经济变量
KPCA
汇率预测
Internet search information
Macroeconomic variables
KPCA
Exchange rate forecast
