GPU并行计算在LSSVM股票预测上的研究与应用  

RESEARCH AND APPLICATIONS OF GPU PARALLEL COMPUTING IN FINANCIAL FORECAST OF LSSVM

作　　者：郑晓薇[1] 侯立斐[1] 于梦玲[1] 

机构地区：[1]辽宁师范大学计算机与信息技术学院,辽宁大连116081

出　　处：《计算机应用与软件》2013年第1期34-36,共3页Computer Applications and Software

基　　金：国家自然科学基金项目(61105059)

摘　　要：金融市场股票的价格预测是投资者们关心的热门问题,随着股票数据规模以及样本维数的日益庞大,人们对于股票预测算法在保证准确之外的预测速度也提出了更高的要求。在最小二乘支持向量机(LSSVM)股票预测算法的基础上,提出一种适宜支持向量机的GPU并行计算模型。实验证明,新方法不仅可以保证预测的精度,而且可以大大缩短预测时间。该方法可以广泛运用到金融领域的大规模数据处理以及预测中,具有较高的应用价值。Stocks price forecast in financial market is the top concern of the investors. Along with the increasingly bulky stock data size and dimensions of sample, people pursue bigber requirement in prediction speed of stock prediction algorithm other than its accuracy. This paper, based on least squares support vector machine （LSSVM） stock prediction algorithm, presents a GPU＇ s parallel computing model suitable for support vector machine. Experiments show that the new method can guarantee the accuracy of the prediction, and can also reduce the prediction time remarkably. This method can be widely applied to large-scale data processing and forecasting in financial sector, and has high applied value.

关 键 词：GPU 并行计算 最小二乘支持向量机 股票预测 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]