Stock Comment Sentiment Analysis Based On Neural Networks And AFA Method

Classical finance takes the validity market as the hypothesis,and believes that the value of financial market is always equal to its basic value,but it cannot explain the'vision'in real life.Behavioral finance emphasizes that market investors' perceptions and emotions will make influence on their decision making,so that causes price fluctuations in the capital market.As an important part of the investment market,the stock market's index changes will also be affected by investor sentiment.Many scholars have used machine learning methods to mine stock market investors' emotions.However,few people have mentioned the problem of unbalanced text classification caused by excessive text noise in real-world application environments.In the process of emotional mining analysis of the stock comments on Eatmoney.com,the paper firstly use SeqGAN algorithm as text generator,oversampling the imbalanced stock comment dataset,and classifying the dataset on the random forest classifier,the method of text oversampling using SeqGAN algorithm is more effective than traditional oversampling methods such as SMOTE,Borderline-SMOTE,and ADASYN,and are more suitable for imbalanced text classification tasks.Secondly for the traditional convolutional neural network can not capture the long-term context in the corpus and the relationship between non-continuous words,this paper introduces the attention mechanism into the classic short-text classification convolutional neural network model,and experiments prove the effectiveness of the improved structure.Finally,the paper uses the AFA method in stochastic fractal theory,combined with constructing investor sentiment index and stock market logarithmic return rate,this paper makes an emotional analysis of stockholders'comments that have been classified,and finds that investor sentiment is highly positively correlated with stock market logarithmic return,and the investors'sentiment possesses long-range correlation.It can provide reference for the subsequent development of appropriate investment strategies.