基于启发式算法的支持向量机选股模型

Stock Selection Model Based on Support Vector Machine within Heuristic Algorithm

发行流通股的上市公司财务数据是高维、复杂的,在利用财务指标对股票进行投资选择时往往难以全面考虑。为了从样本股的大量财务指标中提取出低维、有效的特征信息来构成支持向量机(SVM)的训练集,提出了一种启发式算法(HA)对原始财务数据进行预处理,在保存原始数据特征信息的同时提高了训练精度和训练效率。实证结果中,基于该启发式算法的支持向量机选股模型(HA-SVM)最终构造的股票组合的年收益显著高于同期基准组合的年收益。另外,进一步将被广泛使用于降维和数据特征提取的主成分分析法(PCA)与该启发式算法进行对比分析,结果表明,HA-SVM模型的训练准确率、预测准确率以及所选股票组合的年收益情况均显著高于PCA-SVM模型。

The financial figures of listed company issuing circulation stocks are complicated and high dimensional, which makes it always difficult to make stock selection decisions using the financial figures with all respects considered. To extract the low-dimensional and efficient feature information constituting the training set of support vector machine, this paper proposes a method of heuristic algorithm （HA） to preproeess the original financial figures, which improves the training accuracy and training efficiency along with preservation of the original feature information at the same time. In the empirical results, the stock portfolio constructed by the stock selection model based on support vector machine within heuristic algorithm （HA-SVM） outperforms significantly the A-share index of Shanghai Stock Exchange during the same period. Moreover, this paper makes a comparison analysis between the heuristic algorithm and the principal component analysis （PCA） applied extensively in the dimension reduction and features extraction, and the result shows that the training accuracy, testing accuracy and annual return of the stock portfolio of HA-SVM model are all obviously higher than those of PCA-SVM model.