基于高斯过程回归的上市股价预测模型  

Price Forecasting Method for IPO Based on Gaussian Process Regression

作　　者：杨振舰[1] 夏克文[1] 

机构地区：[1]河北工业大学信息工程学院,天津300401

出　　处：《计算机仿真》2013年第1期293-296,304,共5页Computer Simulation

基　　金：天津市教育科学"十二五"规划课题(HE4068)

摘　　要：在新股上市价格的科学优化预测问题的研究中,由于金融数据复杂,特别是新股价格存在极强的无序性。传统股票价格预测方法只能采用线性变化规律进行准确预测,无法对非线性股票价格进行有效建模,降低股价预测精度。为了提高股票价格预测精度,提出一种高斯过程回归的新股上市价格预测模型,通过提取影响新股上市价格形成的指标因素,用其训练纳斯达克(NASDAQ)新股上市价格的历史数据,以粒子群算法优化高斯过程的超参数来预测新股上市价格。将8家公司的上市股票作为实例进行分析,预测结果表明,高斯过程回归的方法提高股票价格预测精度,能够有效地适用于新股上市价格预测。Scientific price prediction for initial public offerings （IPO） has occupied a very important position in the field of financial investment. The traditional stock price prediction methods can only accurately predict stock prices for linear variation. They are unable to model nonlinear stock prices, which decreases the prediction accuracy of stock prices. In order to improve the prediction accuracy of stock prices, this paper proposed a price forecasting approach for IPO based on Gaussian process regression. It trains the history prices of NASDAQ IPO through extracting the index factors influencing the stock prices. It uses the hyper-parameters of Gaussian process that are optimized by particle swarm optimization （PSO） algorithm to predict IPO price. It analyzes the marketing stocks of eight companies as test samples. The predictive results show that the Gaussian process regression method improves the prediction ac- curacy of stock prices and ts effectively applied to the prices for IPO.

关 键 词：新股上市价格 股价预测 高斯过程回归 纳斯达克 粒子群算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]