基于LSTM的量化股票预测

LSTM Based Quantitative Stock Forecasting

摘要: 股票特征通常夹杂较多噪声数据，而带噪数据会影响股票预测模型的预测精度。本文提出一种对股票数据特征进行量化编码的方法，并使用长短期记忆网络构建预测模型，对量化后的数据进行预测。数据集采用沪深300成分股，在对股票数据量化后进行3分类涨跌幅预测。实验结果表明，使用量化编码对股票特征处理后，预测效果优于使用原始数据预测。

Abstract: The features of stock are usually mixed with many noise data, and noisy data will affect the predic-tion accuracy of stock prediction model. In this paper, a quantitative coding method for stock data features is proposed, and a prediction model is constructed by using short and long term memory network to predict the quantified data. The data set uses the Shanghai and Shenzhen 300 compo-nent stocks, after the stock data quantification carries on the 3 classification rise and fall forecast. The experimental results show that the prediction effect is better than that of the original data after the stock feature is processed by quantitative coding.