Background: The attempt of this research is to propose a novel approach for the efficient prediction of stock prices. The scope of this research extends by including the feature of sentiment analysis using the emotions and opinions carried by social media platforms. The research also analyzes the impact of social media, feeds data and Technical indicators on stock prices for the design of the prediction model.
Objectives: The goal of this research is to analyze and compare the models to predict stock trends by adjusting the feature set.
Methods: The basic technical and new momentum volatility indicators are calculated for the benchmark index values of the stock. The text summarization was applied on collected day-wise tweets for a particular company and then sentiment analysis was performed to get the sentiment value. All these collected features were integrated to form the final dataset and accuracy comparisons were made by experimenting with the algorithms- Support vector machine (SVM), Backpropogation and Long short-term memory (LSTM).
Results: The execution is carried out for each algorithm with 30 epochs. It is observed that the SVM exhibits 2.78%, Backpropogation exhibits 5.02% and LSTM exhibits 10.30 % enhanced performance than the prediction model designed using basic technical indicators. Moreover, along with human sentiment, the SVM provides 5.48%, Backpropogation 5.28% and LSTM 0.07% better accuracy. The standard deviation results are for SVM 1.59, for back propagation 2.46, and LSTM 0.19.
Conclusion: The experimental results show that the standard deviation of LSTM is less than the SVM and back propagation algorithms. Hence, obtaining steady accuracy is highly possible with LSTM.