情感分析與深度學習組態對股價預測的影響研究

At present, most research on stock price prediction is based on technical data.However, with the development of internet social media, investors can easilyreceive information so that they will make different decisions, and thus cause theshort-term changes of the stock price. Therefore, considering both text andtechnical data will be an important direction for improving the accuracy of stockprice forecast.Since the stock price is a time sequence data, this project adopts Transformer tobe our main classifier to predict the stock price. This project will first find out thebest standardized method and configuration of its architecture to improveperformance. And it is utilized as the classifier of sentiment analysis and stockprice predicting models. The best configuration of Transformer can be applied toother research fields that also deal with time sequence data.Secondly, the recent advance in word embedding method has improved theeffectiveness of the sentiment analysis model. In this field, this project willdiscuss the effectiveness of the latest static and dynamic word embeddingmethods to find out the most suitable sentiment analysis model for stock priceprediction.In addition, this project will adopt the best configuration of Transformer with theattention mechanism and the deep learning hybrid model. Both of them havebeen frequently used as a classifier for the sentiment analysis model and stockprice predicting model in recent years. The options of the most suitable classifierare further evaluated.To sum up, this project will first collect text data from three popular social mediaplatforms. The data are trained, tested and verified with the most effectivesentiment analysis model and the most suitable classifier. It will be expected tobuild a state-of-the-art stock price predicting model with high accuracy andapplicability that takes into consideration of both the technical and text data frominternet social media.