Short-term stock trends prediction based on sentiment analysis and machine learning

Investor-generated textual contents have been proved to be the crucial factor that can cause fluctuations in stock price. However, the existing researches only used the equal-weighted method to construct sentiment index for the textual contents, which also rarely considered the impact of financial anomalies. Therefore, in this study, we develop a novel sentiment index to predict the stock trends based on the weighted textual contents and financial anomalies. Specifically, we first propose a novel weighting method to weight each stock review. Then, the day-of-the-week effect and holiday effect are taken into consideration to construct more reliable and realistic modified sentiment index. Experimental results show that the modified sentiment index can effectively improve the predicted ability on stock trends prediction when applying Support Vector Machine (SVM), Decision Tree (DT), Gradient Boosting Decision Tree (GBDT), Random Forest (RF), Naive Bayes (NB), K-Nearest Neighbor (KNN) and Logistic Regression (LR) models.

Automated summary

This study develops a new sentiment index to predict stock trends by considering weighted textual contents and financial anomalies. The experimental results show that the modified sentiment index can effectively improve the prediction ability of stock trends.