| 摘要: |
| [目的]探寻一种有效预测股票价格变化趋势的方法。[方法]在指数平滑异同移动平均线(MACD)指标中加入市场活跃程度(ACT)、波动率(VOL)、离差值(DIF)趋势程度3个指标来构建股票价格变化趋势预测模型。采用热点图对MACD策略中的参数以及股票进行选择、融合,并运用技术分析工具、支持向量机(SVM)与相关向量机(RVM)等机器学习方法对MACD策略中产生的交易信号进行优化,筛选特征变量。[结果]将设计的股票价格变化趋势预测策略的数据换成A股全市场数据并进行回测,发现近10年的年化收益率(14.8%)胜过沪深300指数(7.201%),而且使用A股全市场的数据有效避免了幸存者的偏差。[结论]优化改进的股票价格变化趋势预测模型可以在一定程度上预测股票上升的趋势,有效规避风险。 |
| 关键词: 股票价格 量化交易 机器学习 |
| DOI：10.13657/j.cnki.gxkxyxb.20170320.001 |
| 投稿时间：2017-01-10 |
| 基金项目:广西科学研究与技术开发计划项目(桂科能1140008-3B)资助。 |
|
| Predicting Stock Price Trend by Optimizing MACD Model |
|
YANG Yongjie
|
| (School of Computer, Electronics and Information in Guangxi University, Nanning, Guangxi, 530004, China) |
| Abstract: |
| [Objective] Explore a way to effectively predict stock price trends.[Methods] Three indices of market activity,volatility,and DIF trend are added into the Moving Average Convergence and Divergence (MACD) strategy to generate stock price trend prediction model.The parameters of MACD and a subset of stocks are selected by the heat map.Technical analysis and machine learning methods such as support vector machine and relevance vector machine are used to improve the signaling capability of the MACD strategies.[Results] We obtain 14.8% annualized out of sample back analysis return in Chinese A-share market for the period of 1/4/2006 to 8/31/2016,which is significantly higher than Hushen300 annual return of 7.201% for the same period. Survivorship bias is avoided with all stocks in Chinese A-share market.[Conclusion] Our results demonstrate that to a certain extent the improved stock price trend prediction model can capture the upward tendency of Chinese stock market and effectively reduce risks. |
| Key words: stock price trending quantitative trading machine learning |
