MATEC Web Conf.
Volume 381, 20231st International Conference on Modern Technologies in Mechanical & Materials Engineering (MTME-2023)
|Number of page(s)||11|
|Published online||13 June 2023|
A hybrid sentiment based stock price prediction model using machine learning
Department of Industrial Engineering, University of Engineering and Technology Taxila, Punjab 47050, Pakistan
* Corresponding author: email@example.com
Accurate stock market prediction is highly desirable to corporations and investors. In this study a deep learning model based on LSTM, BiLSTM with attention mechanism used to predict stocks closing price for next 30 days of two banks listed in Pakistan Stock Exchange. For accurate stock price prediction, it is necessary to consider volatile factors such as news sentiments along with historical data. This study covers that aspect by incorporating news sentiments along with historical stock data that is distributed over a span of ten years from Jan 2011 to July 2021. Preprocessing and sentiment analysis of data was performed using python NLTK module. After that we built a univariate deep learning model based on four layers of LSTM and one dense layer to combine all layers and performed a prediction on train and test data followed by a multivariate deep learning model based on BiLSTM with self-attention mechanism and found out that incorporation of news sentiments really improved the prediction accuracy by reducing the values of mean squared error. Finally, we did the prediction for next 30 days of stock closing price of two banks and compared those predicted prices with actual prices and got quite accurate results.
Key words: Deep Learning / Stock price prediction / News Sentiments / LSTM / BiLSTM
© The Authors, published by EDP Sciences, 2023
This is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.