A novel deep learning model for stock market prediction using a sentiment analysis system from authoritative financial website’s data

Abstract

Listen

Translate article

The use of deep learning, specifically time series neural networks, in predicting stock market trends has emerged as a significant use case in financial analysis. However, the complex interrelationships and instability of the stock market have made the timely and accurate prediction of its behaviour as a confronting endeavour. To address this difficulty, in this research work a stock market index prediction model called SenT-In, which combines the with a sentiment awareness model. A sentiment awareness model using Convolutional Neural Networks (CNN) and Gated Recurrent Unit (GRU) is proposed to calculate the sentiment index of a large volume of news articles collected from reputable financial websites. In addition, a sentiment attention method is developed to combine stock data and news sentiment index as the input for training and predicting using the SenT-In network, which is both simple and efficient. The proposed model is evaluated in four different stock market datasets which include FSTE, SSE, Nifty 50 and S&P 500. On comparing the results with conventional deep learning algorithms such as GRU, LSTM, CNN and SVM, proposed SenT-In outperforms existing methods in accuracy with 9%, F1-Score with 7%, AUC-ROC curve with 13% and PR-AUC curve with 9% efficiency (on average).