Abstract:
"Stock price prediction Is an important area of research in the field of financial time series
prediction. Most existing approaches to this problem use past historical data of stocks to
derive a prediction for the future value of a stock. But latest research indicates that including sentiments of related financial news articles may increase the accuracy of such predictions. A major issue when using news articles for the prediction is the impact of fake news that can result in inaccurate predictions as observed by researchers.
This research aims to answer the issue of fake news detection in such systems by integrating a fake news filter to filter news articles before they are used for the sentiment detection process. This was achieved by training a fake news classifier on the ISOT dataset. Another research gap detected was the use of novel combinations of hybrid deep learning algorithms. Research has indicated that hybrid algorithms are more accurate than traditional or deep learning ml models therefore a novel combination of ARFIMA-BILSTM was introduced in this research. Tests indicate that the AFRIMA-BILSTM algorithm achieves a lower RMSE value compared to the ARIMA-LSTM algorithm which indicates the superiority of this novel approach."
