【摘要】 由于股票数据存在噪声，传统的机器学习模型并不能很好地预测股票的涨跌。为了帮助投资者了解金融市场的发展趋势，构造有效的投资组合从而赢得超额收益，将对噪声数据具有良好鲁棒性的TWSVM算法应用到量化投资中。构造不同分类数据，以预测的正确率作为评价指标对TWSVM的核函数进行评估，发现poly核函数在TWSVM算法中具有稳定性。建立基于TWSVM的量化投资策略，以上证50股选股实验为例模拟交易并与RF、SVM、Logistic算法对比说明策略的有效性，实验结果表明，基于TWSVM量化投资策略的年化收益率比其他三者提高约3百分点至11百分点。更多还原

【Abstract】 Because of the noise of stock data, the traditional machine learning model cannot predict the rise and fall of stocks well. In order to help investors understand the development trend of the financial market and construct an effective portfolio to win excess returns, the twin support vector machines(TWSVM) algorithm with good robustness for noise data is applied to quantitative investment. Different classification data were constructed, and the prediction accuracy was used as the evaluation index to evaluate the kernel function of TWSVM. It was found that the poly kernel function was stable in TWSVM algorithm. A quantitative investment strategy based on TWSVM was established. Taking stock selection experiment of SSE 50 as an example, the trading was simulated and compared with RF, SVM and Logistic algorithm to illustrate the effectiveness of the strategy. The experimental results show that the annualized return rate of the quantitative investment strategy based on TWSVM is 3～11 percentage points higher than that of the other three.更多还原