基于两类核函数的TSVR在股价预测中的比较  

Comparison of TSVR Based on Two Kinds of Kernel Functions in Stock Price Forecasting

作　　者：尹湘锋[1] 崔浩锋 文雪婷 Yin Xiangfeng;Cui Haofeng;Wen Xueting(School of Mathematics and Computational Science,Hunan University of Science and Technology,Xiangtan Hunan 411201,China)

机构地区：[1]湖南科技大学数学与计算科学学院,湖南湘潭411201

出　　处：《统计与决策》2021年第12期43-46,共4页Statistics & Decision

摘　　要：股价预测是一个复杂的非线性问题,随着数据分析技术的不断发展,应用机器学习方法来进行股票量化分析的研究也越来越多。文章用非线性孪生支持向量回归对股票价格的走势进行分析,对股票价格进行预测。在应用中分别对使用线性核函数和多项式核函数的非线性孪生支持向量回归预测性能进行了比较;在回归中的参数寻优上,分别利用遗传算法和粒子群优化算法进行寻优。结果表明:线性核孪生支持向量回归和多项式核孪生支持向量回归对股价走势的分析及价格预测均可提供一定程度的参考,其中多项式核孪生支持向量回归有较高的精度;参数寻优上,遗传算法优化效果较好。Stock price forecasting is a complex nonlinear issue.With the continuous development of data analysis technology,more and more studies apply machine learning methods to stock quantitative analysis.This paper uses the nonlinear twin support vector regression to analyze the trend of stock price and predict the stock price.The forecasting performance of nonlinear twin support vector regression with linear kernel function and polynomial kernel function is compared respectively in application;in the parameter optimization of regression are used genetic algorithm and particle swarm optimization algorithm.The results show that the linear kernel twin support vector regression and polynomial kernel twin support vector regression can provide some reference for stock price trend analysis and price prediction,and polynomial kernel twin support vector regression has higher accuracy;and that in parameter optimization,genetic algorithm has better optimization effect.

关 键 词：孪生支持向量机 线性核 多项式核 GA PSO 

分 类 号：O21[理学—概率论与数理统计]