基于机器学习模型预测A股市场高送转股票

Prediction of Highly Convertible Stocks in A-share Market Based on Machine Learning Model

A股高送转作为我国政策特有的现象,对其精确、有依据的预测有一定的研究价值.本文对A股数据集进行缺失值、异常值、标准化等数据处理,采取特征选择中过滤法、包裹法、嵌入法等方法提取特征,结合经济学意义得到最终特征.数据的不平衡采取过采样、欠采样以及SMOTE采样等方法处理.最后采用基于Stacking算法融合模型,第1层学习器采用5种分类机器学习模型,并进行超参数调整;第2层采用LightGBM模型预测,以F1分数为评价指标,进行对A股市场预测哪些公司可能会实施高送转.

As a special phenomenon of our country′s policy,the high transfer of A-share has certain research value for its accurate and reasonable prediction.In this paper,the A-share data set is processed with missing value,abnormal value,and standardization,and the features are extracted by filtering,wrapping,and embedded methods in feature selection,and the final features are obtained by combining the economic significance.Data imbalance is processed by methods such as over-sampling,under-sampling and SMOTE sampling.Finally,a fusion model based on the stacking algorithm is used.In the first layer,5 classification machine learning models are used,and the hyperparameters are adjusted.In the second layer,LightGBM model,along with the F1 score as evaluation indicators,are used to predict which companies in the Ashare market may implement high conversion.