| 应用罚函数方法构建广义指数因子预报模型——黄金价格预测的实证分析 |
| |
| 作者姓名： | 严威 尹伟 缪柏其 叶五一 |
| |
| 作者单位： | 中国科学技术大学统计与金融系, 合肥 230026 |
| |
| 基金项目： | 国家自然科学基金青年科学基金(71001095)和高等学校博士学科点专项科研基金(20103402120010)资助 |
| |
| 摘 要： | 采用不同的损失函数和罚函数构建了广义指数预报因子模型,用该模型来预测国际黄金价格. 构建方法包括:1)岭估计方法;2) 基于L1、L2以及二者结合的损失函数LM,利用LASSO和SCAD 2种罚函数选取不同参数EWMA的线性组合作为预报因子. 实证检验表明,该方法构建的模型有效改进了单参数EWMA预测模型,其预测精度优于已有方法.
|
| 关 键 词： | 广义指数因子 黄金价格预测 变量选择 罚函数 损失函数 |
| 收稿时间： | 2010-11-15 |
| 修稿时间： | 2011-03-24 |
Constructing generalized exponential predictors via penalty methods: empirical analysis on gold price |
| |
| Authors: | YAN Wei YIN Wei MIAO Bai-Qi YE Wu-Yi |
| |
| Institution: | Department of Statistics & Finance, University of Science & Technology of China, Hefei 230026, China |
| |
| Abstract: | We construct generalized exponential predictors for forecasting gold price using different loss and penalty functions. The construction methods include: 1) ridge regression and 2) selection of linear combinations of EWMA predictors with different parameters by adding LASSO and SCAD penalties based on L1,L2 and the LM loss function which combines both L1 and L2. Practical data show that our models improve the single parameter EWMA model effectively and they perform better than the models suggested in the literature. |
| |
| Keywords: | generalized exponential predictors gold price forecast variable selection penalty functions loss function |
|
| 点击此处可从《》浏览原始摘要信息 |
| 点击此处可从《》下载免费的PDF全文 |
