| 基于线性或非线性结构自动识别的股价预测 |
| Stock price prediction based on automatic identification of linear and nonlinear structures |
| 摘要点击 29 全文点击 0 投稿时间：2022-03-17 修订日期：2024-02-28 |
| 查看/发表评论 下载PDF阅读器 |
| 中文关键词 股价预测; Lasso方法; 群Lasso方法; 加性模型; LANEM方法 |
| 英文关键词 stock price forecast; Lasso; Group Lasso; the additive model; LANEM |
| 基金项目 无 |
| 作者 | 单位 | 邮编 | | 梁焙婷 | 暨南大学经济学院统计学系 | 510632 | | 王斌会 | 暨南大学经济管理国家实验教学中心 | | | 王国长 | 暨南大学经济学院统计学系 | | | 庞檬缘 | 新加坡国立大学理学院 | |
|
| 中文摘要 |
| 针对加性模型中预测变量线性和非线性结构的自动识别问题, 提出了一种基于Lasso和群Lasso方法的变量选择和参数估计方法LANEM. 首先结合Lasso惩罚和群Lasso惩罚, 通过使用三次自然样条基底同时选出无关变量、线性变量以及非线性变量, 并利用上证指数数据, 采用计算机仿真模拟方法进行股价预测. 结果表明, LANEM方法基于股票数据能自动识别出线性预测变量、非线性预测变量和无关预测变量, 且结合最小二乘方法得到的LANEMLS方法具有最小的预测误差. 同时, 稳健性检验证明了LANEM方法不受股票指标、类型和时间的影响, 应用股票场景广泛. 相较于Lasso 和群Lasso方法, LANEM方法有效提高股票指数预测的准确率, 对股票的预测和趋势性研究具有现实意义. |
| 英文摘要 |
| This paper studies the automatic identification of linear and nonlinear structures of predictors in the additive model. A method LANEM based on Lasso method and group Lasso method has been proposed, which could simultaneously perform variable selection and parameter estimation. Firstly, the LANEM method combines with Lasso penalty and group Lasso penalty, and uses a cubic natural spline basis to select linear predictors, nonlinear predictors and uncorrelated predictors simultaneously. Moreover, the LANEM method uses Shanghai Composite Index data and computer simulation methods for stock price prediction. The results indicate that the LANEM method can automatically identify linear predictors, nonlinear predictors and uncorrelated predictors based on stock data, and the LANEMLS method combined with the least squares method has the smallest prediction error. Meanwhile, the robustness test proves that the LANEM method is independent of stock index, type and time, and can be applied to a wide range of stock scenarios. Compared with the Lasso and group Lasso methods, the LANEM method can effectively improve the accuracy of stock index prediction, and have practical significance for the research of stock prediction and trend. |
| 关闭 |
|
|
|
|
|
