基于SVM的黄金价格预测模型及其参数优化

发布时间：2018-06-07 18:09

本文选题：黄金价格 + SVM模型　；参考：《山东大学》2017年硕士论文

【摘要】：2016年以来伴随着全球经济的复苏大宗商品结束了连续5年的熊市,市场交易量逐渐增大,市场逐渐活跃,重新回到投资者的配置资产中。CTA策略近两年的不凡表现以及机器学习的崛起,让研究者、投资者对机器学习应用到大宗商品交易产生了很大的兴趣。本文基于机器学习中的支持向量机模型,选取大宗商品中金融属性最强的黄金,对黄金的价格进行预测,并初步尝试构建一个择时策略。本文共分为五个部分:第一部分是绪论部分,描述了支持向量机预测黄金价格模型的实际意义以及整个文章研究的方法和论述侧重点;第二部分,对黄金品种的属性和功能进行了阐述,基于此分析了影响黄金价格的主要因素,最后结合实际意义、相关性分析以及格兰杰因果检验选取了其中7个影响因素作为支持向量机建模用的特征因子。本文的第三部分是文章核心,首先介绍了 SVM算法的基本原理及应用,并给出了实现算法流程,然后介绍了优化SVM参数的网格搜索法和粒子群优化算法的基本原理及实现算法,第三章基于MATLAB基于上述原理算法对黄金价格进行了预测,并以最小化MSE为目标优化了 SVM模型中的惩罚项参数以及核函数中的参数;第四章基于第三章预测黄金价格的SVM模型,初步构建了一个择时交易策略,并给出在回测样本中的回测结果。第五章,对文中模型的不足和创新点进行了总结,最后结合全文对模型以及优化结果进行了简要评价,对后续进一步研究做了展望和指明了方向。本文中涉及的数据均来自Bloomberg客户端,数据处理和模型实现基于MATLAB 软件。
[Abstract]:The commodities recovery that has accompanied the global economy since 2016 ended five consecutive years of bear markets, with the volume of trading increasing and the market becoming more active. The return to investors' allocation of assets. CTA strategy in the past two years, and the rise of machine learning, so that researchers, investors have a lot of interest in the application of machine learning in commodity trading. Based on the support vector machine model in machine learning, this paper selects gold, which has the strongest financial attributes in commodities, to predict the price of gold, and tries to construct a timing strategy. This paper is divided into five parts: the first part is the introduction part, which describes the practical significance of the support vector machine forecasting gold price model, the research methods and the emphases of the whole article; the second part, This paper expounds the attribute and function of gold variety, analyzes the main factors influencing gold price based on this, and finally combines with the practical significance. Correlation analysis and Granger causality test selected 7 of them as feature factors for SVM modeling. The third part of this paper is the core of the article. Firstly, the basic principle and application of SVM algorithm are introduced, and the realization flow is given. Then, the basic principle and realization algorithm of mesh search method and particle swarm optimization algorithm for optimizing SVM parameters are introduced. In chapter 3, the gold price is predicted based on MATLAB, and the penalty parameters and kernel parameters in SVM model are optimized with the goal of minimizing MSE. Chapter 4 is based on the SVM model for predicting gold price in chapter 3. A timing trading strategy is preliminarily constructed, and the results in the back test sample are given. In the fifth chapter, the shortcomings and innovations of the model are summarized. Finally, the model and optimization results are briefly evaluated, and the further research is prospected and the direction is pointed out. The data involved in this paper come from Bloomberg client, data processing and model implementation are based on MATLAB software.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;F830.94

【相似文献】