基于时间序列模型和情感分析的情感趋势预测

Predicting sentiment trend by time series model and sentiment analysis

为解决传统情感分析方法无法对公众未来情感走势变化有效预测的问题,提出一种将时间序列模型与情感分析相结合的情感趋势预测方法。采用深度学习模型对股市论坛实时评论信息进行情感分类,统计固定时间单位的情感值,构建情感值时间序列,提出ARIMA-GARCH时间序列模型,对情感值时间序列进行建模分析,预测投资者的情感走势。实验结果表明,该方法对于情感趋势的预测结果合理,误差较小。同时,发现投资者情感趋势与股市涨跌幅走势相似,为投资决策提供了参考。

To solve the problem that traditional sentiment analysis methods cannot effectively predict the future changes of public sentiment trend,a method of sentiment trend prediction which combined time series model with sentiment analysis was proposed.Deep learning model was used to classify the sentiment of stock market forum reviews data,and sentiment value over fixed time unit was calculated to form the sentiment value time series.The ARIMA-GARCH model was formulated to analyze the sentiment value time series and predict the sentiment trend of investors.Computational results show that the method proposed is of rational predicting result with less error and satisfied effect.Furthermore,it can be found that investor sentiment trend is similar to the advance-decline trend of stock market,providing a reference for future investment decision-making.