结合投资者情绪运用LSTM-GAN模型预测股票

Combining Investor Sentiment, Using LSTM-GAN Model to Forecast Stocks

摘要: 本研究运用股票市场历史数据以及投资者情绪数据，通过实证方法和对比分析方法，探讨了长短期记忆网络模型和对抗生成网络模型以及二者结合在股票收盘价预测中的效果。通过机器学习模型对这些数据进行分析，实验结果表明，混合模型相比于单独的长短期记忆网络模型和对抗生成网络模型误差更低，效果更好，并且模型中加入投资者情绪后，显著提高了股票预测的准确性和鲁棒性。

Abstract: This study utilizes historical data of the stock market and investor sentiment data. Through empirical methods and comparative analysis methods, it explores the effects of the long short-term memory network model and the adversarial generative network model, as well as their combination, in the prediction of stock closing prices. These data were analyzed through the machine learning model. The experimental results show that the hybrid model has lower errors and better effects compared with the individual long short-term memory network model and the adversarial generative network model. Moreover, after adding investor sentiment to the model, the accuracy and robustness of stock prediction were significantly improved.