基于社交媒体文本信息的金融时序预测  

Financial time series prediction based on social media text information

作　　者：李大舟 于沛 高巍 马辉 LI Da-zhou;YU Pei;GAO Wei;MA Hui(School of Computer Science and Technology,Shenyang University of Chemical Technology,Shenyang 110142,China)

机构地区：[1]沈阳化工大学计算机科学与技术学院,辽宁沈阳110142

出　　处：《计算机工程与设计》2021年第8期2224-2231,共8页Computer Engineering and Design

基　　金：辽宁省教育厅科学技术研究基金项目(L2016011);辽宁省教育厅科学研究基金项目(LQ2017008);辽宁省博士启动基金项目(201601196)。

摘　　要：针对传统股票趋势预测模型中忽略社交媒体文本信息对股价变化的影响和时间序列的平稳性处理、长期依赖等问题,提出一种融合社交媒体文本信息和LSTM的股票趋势预测模型(BiTCN-LSTM)。该模型分为情感分析和金融时序预测两部分。情感分析层将社交媒体文本信息输入到双向时间卷积网络进行特征提取和情感分析,得到积极或者消极的情感分类表示;金融时序预测层使用LSTM神经网络,将差分运算后的股票历史数据和文本情感特征向量加权融合作为网络输入,完成金融时序预测任务。通过上海证券综合指数数据集的实验验证,与传统金融时序预测模型相比,该模型的RMSE指标降低3.44-43.62。To solve the problems of ignoring the influence of social media text information on stock price change and the stationary processing and long-term dependence of time series in the traditional stock trend prediction model,a stock trend prediction model(BiTCN-LSTM)integrating social media text information and LSTM was proposed.The model was divided into two parts including emotion analysis and financial timing prediction.The emotion analysis layer of this model inputted the text information of social media into the two-way time convolutional network for feature extraction and emotion analysis to obtain the classified representation of positive or negative emotions.The financial timing prediction layer of this model used LSTM neural network,and took the weighted fusion of stock historical data and text emotional feature vector after difference operation as the network input to complete the financial timing prediction task.Experimental results of Shanghai Securities Composite index data set show that this model has better prediction effect than the traditional financial timing forecasting model.

关 键 词：情感分析 双向时间卷积网络 差分运算 长短时记忆 金融时间序列预测 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]