Proceedings of The 12th International Conference on Management, Economics and Humanities
Prediction of stock price direction using machine learning models: based on sentiment analysis
Chengjiao Li, Sarah Fores
Prediction of the stock trend is always a challenging but attractive topic. In this article, apart from widely used financial indicators such as MACD, the investors’ sentiments are also considered in models to predict the movement of SSE composite index. In addition, sector vector machine (SVM) and artificial neural network (ANN) are combined with genetic algorithm (GA) to predict. GA is used to optimize machine learning models. The results show that considering investor sentiments could only remain the prediction performance at the same level at least, and even make it worse. The sentiments extracted under different training data have a relatively large difference. Moreover, when considering financial indicators only, SVM performs as well as SVM-GA and performs better than ANN and ANN-GA. GA doesn’t improve the performance. When considering financial indicators and investor sentiments together, SVM-GA performs the best and ANN-GA the worse, but ANN outperforms SVM. GA improves the performance of SVM but decreases that of ANN.
keywords: Artificial neural network; Genetic algorithm; Prediction of stock price direction; Section vector machine; Sentiment analysis.