基于改进GM（1，N）和优化SVM组合模型的股票价格预测

发布时间：2018-01-07 15:16

本文关键词：基于改进GM（1，N）和优化SVM组合模型的股票价格预测　出处：《哈尔滨工业大学》2013年硕士论文　论文类型：学位论文

【摘要】：股票市场是资本市场的重要组成部分，也被看作是国民经济的“晴雨表”，在优化资金配置、筹资融资以及资产保值增值等领域扮演着重要的角色。但股票市场风险与收益并存，，投资风险较大，能够使用一些方法和相关模型分析金融市场的走势，对股票价格进行估计和预测，为投资者提供投资决策参考，具有极为重要的现实意义。本文通过文献的回顾和总结，首先从宏观和微观两个角度分析了影响股票价格的因素，初步选取了GDP、利率、汇率等八个指标作为股票价格预测的输入变量，为精简输入变量个数，提高预测精度，使用了与逐步回归、相关分析相结合的改进灰色关联分析方法对影响股票价格因素的变量进行了二次特征提取，最终选取了M2、汇率、企业景气指数、标准普尔500指数四个变量作为预测模型的输入变量。在股票价格预测上，构建了基于诱导有序加权平均算子的组合模型，组合模型包括改进的GM（1，N）模型和不确定知识粒子群（PSO-UK）优化支持向量机模型，两个模型对股票价格的线性和非线性部分进行分别预测，并使用诱导有序加权平均算子确定两个模型的权重，前文确定的四个变量作为输入值，上证指数作为输出值，对模型进行了实证检验。使用平均绝对百分比误差、均方百分比误差等指标，与原始GM（1，1）模型、原始GM（1，N）模型、粒子群优化支持向量机模型以及BP神经网络模型进行对比分析，对比结果表示文中提出的组合模型具有更优的推广和预测能力，验证了组合模型的优越性，组合模型能够为投资者对股市投资决策提供参考依据。
[Abstract]:The stock market is an important part of the capital market, is also regarded as the "barometer" of the national economy, in the optimization of the allocation of funds. Financing and asset preservation and appreciation play an important role, but the stock market risk and return co-exist, investment risk is high, can use some methods and related models to analyze the trend of financial markets. It is of great practical significance to estimate and forecast the stock price and to provide investors with reference for investment decision. This paper reviews and summarizes the literature, first from the macro and micro perspective to analyze the impact of the stock price factors, the preliminary selection of GDP, interest rate. Exchange rate and other eight indexes are used as input variables of stock price prediction. In order to reduce the number of input variables and improve the accuracy of prediction, we use and step by step regression. The improved grey correlation analysis method combined with correlation analysis extracts the quadratic feature of the variables that affect the factors of stock price, and finally selects M2, exchange rate, business climate index. The four variables of the S & P 500 index are used as input variables of the forecasting model. In the prediction of stock price, a combination model based on induced ordered weighted average operator is constructed. The combination model includes the improved GM(1. N) model and PSO-UK (uncertain knowledge Particle Swarm Optimization) optimization support vector machine model. The two models predict the linear and nonlinear parts of stock price separately. Using the induced ordered weighted average operator to determine the weights of the two models, the four variables determined in the previous paper as the input value, and the Shanghai stock index as the output value, the model is tested empirically. The average absolute percentage error, mean square percentage error and so on were compared with the original GM1 / 1) model and the original GMU 1 / N) model. The particle swarm optimization support vector machine model and BP neural network model are compared and analyzed. The comparison results show that the combination model proposed in this paper has better generalization and prediction ability, which verifies the superiority of the combined model. The portfolio model can provide a reference for investors to make decisions on stock market.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F830.91;F224

【相似文献】