证券投资管理中收益率的预测方法研究

发布时间：2018-03-29 12:22

本文选题：证券投资　切入点：回归方法　出处：《中南大学》2013年硕士论文

【摘要】：证券投资活动是现代经济市场中一个有效和最常见的投资行为,投资收益率是所有投资者最为关心的决策依据。未来收益率的非确定性对证券投资管理中收益率的预测有着重要的作用。本文将采用三种预测算法：回归预测算法,指数平滑预测算法和GM(1,1)预测算法,对证券收益率进行预测,并对这些方法加以比较,选择最合适的方法。实证数值结果表明： 1、对原始信息较多,自变量和因变量之间相关性较强,只是做短期预测时,多元线性回归方法的预测效果好。 2、当时间序列变化不大比较稳定时,并原始数据资源较少,采用指数平滑方法的预测效果更佳。 3、当原始信息较少,无规律,光滑离散原始序列且不服从典型分布时,选用GM(1,1)方法得到的预测效果最佳。本文逐一详细介绍了指数平滑预测算法、多元线性回归预测算法和GM(1,1)预测算法,就三种算法的基本原理、建模过程以及误差的检验进行了一一阐述,并通过经济市场历史数据对三种算法进行了实证效果比较,并对其预测效果、使用条件及特点进行了分析,为经营者决策提供实证的科学依据。
[Abstract]:Securities investment is an effective and most common investment behavior in modern economic market. The return on investment is the most concerned decision basis for all investors. The uncertainty of future returns plays an important role in the prediction of returns in securities investment management. This paper will adopt three forecasting algorithms: regression prediction algorithm. Exponential smoothing prediction algorithm and GM1 / 1) forecasting algorithm are used to predict the return rate of securities. The comparison of these methods and the selection of the most appropriate method. The empirical numerical results show that:. 1. For the original information, the correlation between independent variables and dependent variables is strong, but when short-term prediction is made, the prediction effect of multivariate linear regression method is good. 2, when the time series change is not stable, and the raw data resources are less, the prediction effect of exponential smoothing method is better. 3. When the original information is less, irregular, smooth discrete original sequence and dissatisfied from the typical distribution, the best prediction effect is obtained by using the GM1 (1) method. In this paper, the exponential smoothing prediction algorithm, multivariate linear regression prediction algorithm and GM-1) prediction algorithm are introduced in detail. The basic principles of the three algorithms, the modeling process and the error test are described one by one. Through the historical data of the economic market, the empirical results of the three algorithms are compared, and the prediction effect, using conditions and characteristics are analyzed, which provides the scientific basis for the decision of the operator.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F830.91;F224

【参考文献】