ML-TEA:一套基于机器学习和技术分析的量化投资算法

发布时间：2018-03-19 15:58

本文选题：量化投资　切入点：机器学习　出处：《系统工程理论与实践》2017年05期 　论文类型：期刊论文

【摘要】：量化投资尝试利用计算机算法来预测证券的价格和进行证券的交易,并从中获取超额收益,是系统工程在金融投资领域的重要应用.本文设计了一套基于机器学习和技术指标的量化投资算法ML-TEA(machine learning and technical analysis).该模型以技术指标作为输入变量,再分别通过不同的机器学习算法来预测股票数日之后的涨跌方向,并根据预测的方向来构建投资组合.实证结果显示:第一,三种模型的年化收益率都在25%以上,远超大盘指数的10.60%、买入持有策略的3%以及现有策略.从风险调节绩效(夏普比率、特雷纳比率和詹森绩效)来看,三种策略也都远超基准策略和现有策略.以夏普比率为例,三种策略均在1.50以上,而市场指数的夏普比率为0.38.第二,Ada-TEA和SVM-TEA都可以容忍远高于市场实际成本的交易成本.
[Abstract]:Quantitative investment attempts to use computer algorithms to predict the price of securities and trade them, and to earn excess returns from them. It is an important application of system engineering in the field of financial investment. In this paper, a set of quantitative investment algorithm ML-TEA(machine learning and technical analysis based on machine learning and technical index is designed. Then we use different machine learning algorithms to predict the direction of stock's rise and fall after the vote day, and construct the investment portfolio according to the forecast direction. The empirical results show that: first, the annual return rate of the three models is above 25%. Far more than 10.60 percent of the market index, 3% of the buying and holding strategy, and the existing strategy. In terms of risk-adjusted performance (Sharp ratio, Trayner ratio, and Jason performance), all three strategies also far exceed the benchmark strategy and the existing strategy. Take Sharp ratio as an example. All three strategies are above 1. 50, while the Sharp ratio of the market index is 0. 38. Both Ada-TEA and SVM-TEA can tolerate transaction costs that are much higher than the actual market costs.
【作者单位】：武汉大学经济与管理学院金融系;康奈尔大学运筹与信息工程学院;
【基金】：国家自然科学基金(71401128,91646206) 教育部留学回国人员科研启动基金武汉大学人文社科青年学者学术团队项目(16WSKTD008)~~
【分类号】：F832.51

【相似文献】