基于组合模型的隔夜Shibor预测 Alternative Title OVERNIGHT SHIBOR FORECASTING BASED ON COMBINATION MODELS 李敏 Thesis Advisor 严定琪 2018-03-20 Degree Grantor 兰州大学 Place of Conferral 兰州 Degree Name 硕士 Keyword Shibor 时间序列 神经网络 组合预测 集合经验模态分解 Abstract 上海银行间同业拆放利率，是金融同业之间互相拆借头寸进行融资的利率，具备公开性和市场化的特点，能实时精确的反应出货币市场甚至整个金融市场上较短时间内的资金供求关系，对于中央银行货币政策调控有指导意义。 本文选取隔夜Shibor值作为预测研究的对象，同时建立了两种组合模型。一是对数据做EEMD分解，对分解后的数据建立BP-Adaboost预测；一是针对单纯使用时间序列模型的线性局限性，以及神经网络模型易于陷入局部最优解的不足，建立了一个将时间序列模型的线性预测结果和BP非线性预测结果进行组合的ARIMA-BP-Adaboost模型。使用ARIMA模型对线性部分进行预测，所得的残差作为非线性预测部分的期望输出，采用BP-Adaboost对原始数据和残差值进行训练，输出预测残差，并将其和ARIMA模型的预测值相加，即得到ARIMA-BP-Adaboost组合模型的预测结果。本文选取多种模型对隔夜Shibor数据进行了预测，通过对各模型的拟合程度和预测误差进行比较，提出了两种组合模型预测方法，实证分析表明组合模型的预测效果要明显好于单纯的神经网络预测模型和时间序列预测模型。 Other Abstract Shibor,the full name is Shanghai Interbank Offered Rate, which is the interest rate that financial peers use to borrow money from each other for financing, it has the characteristics of being openness and marketization. Shibor can reflect the short-term capital supply and demand relationships timely and accurately in the money market and even the entire financial market. It has guiding significance for the control of monetary policies of central banks in various countries. In this paper, the overnight Shibor value is selected as the object of the prediction study, and two combination models are established respectively. One is to do EEMD decomposition of the data and establish BP-Adaboost predictions for the decomposed data; the other is to address the linear limitations of simply using the time series model and the shortcomings of neural network models that are prone to fall into local optimal solutions. The ARIMA-BP-Adaboost model is a predictive model that combines the linear prediction results of the time series model with the nonlinear prediction results of the neural network model. The ARIMA model is used to predict the linear part. The residual is used as the expected output of the nonlinear prediction part. The original data and residual values are trained by BP-Adaboost. The prediction residuals are output, and the residuals of the predicted outputs are summed. The prediction results of the ARIMA model are superimposed, and the prediction results of the ARIMA-BP-Adaboost combination model are obtained. In this paper, we use a variety of models to forecast overnight Shibor data, and finally put forward two combined model prediction methods. By comparing the degree of fit of each model and the prediction error, we find that the prediction accuracy of the combined model is significantly better than simply using Neural network prediction model and time series prediction model. URL 查看原文 Language 中文 Document Type 学位论文 Identifier https://ir.lzu.edu.cn/handle/262010/225066 Collection 数学与统计学院 Recommended CitationGB/T 7714 李敏. 基于组合模型的隔夜Shibor预测[D]. 兰州. 兰州大学,2018.
 Files in This Item: There are no files associated with this item.
 Related Services Recommend this item Bookmark Usage statistics Export to Endnote Altmetrics Score Google Scholar Similar articles in Google Scholar [李敏]'s Articles Baidu academic Similar articles in Baidu academic [李敏]'s Articles Bing Scholar Similar articles in Bing Scholar [李敏]'s Articles Terms of Use No data! Social Bookmark/Share
No comment.