Abstract: To better study the stock index prediction problem, a stock index forecasting method based on feature selection and LSTM model is proposed and the predicting ability is improved through selection of feature parameters. The method includes three steps, selecting feature parameters, applying system clustering method and performing principal component analysis to to reduce the dimension. In experiments, the LSTM model is applied to predicte Nasdaq index and s&p 500 index. The experimental results show that the computation load of the proposed method is light and the prediction speed and accuracy are improved.

Key words: stock index forecast, feature selection, system clustering, Principal Component Analysis（PCA）

摘要： 为了更好地研究股指预测问题，提出了基于特征选取与LSTM模型的股指预测方法，该方法从优化特征参数选取角度对模型预测能力进行提升，包含全面选取特征参数、应用系统聚类法进行特征分类、应用主成分分析对分类特征进行降维三个步骤。在实证论证中，应用LSTM模型对纳斯达克股票指数数据和标普500指数数据进行预测，实验结果表明所提出的方法计算量小，预测结果在速度和准确度两方面分析均得到显著提升。

关键词: 关键词：股指预测, 特征选取, 系统聚类, 主成分分析