Opinion mining for stock trend prediction using deep learning

Stock market prediction by using Machine Learning (ML) models has been a hot topic of research for more than a decade. Combined with the power of sentiment analysis and ML, social media posts like tweets and financial news can have a big impact on how accurately stock market price predictions are made. Our proposed work makes use of social media posts from Twitter and Google Finance data to predict the price of the stock market using DL techniques. To perform this job, we propose a method that first introduces a concept of fake tweets detection and removal using BERT and then we prepare our dataset using preprocessing techniques. The feature extraction is performed using FastText and VADER, after that the prediction is made through the Deep Belief Network. Two datasets (i.e., Sentiment140 and Twitter real-time data) were used in the experiments and attained 89.12% accuracy along with precision, recall, and F1-Score of 90.15%, 89.12%, and 89.67%. This research not only refines stock market prediction accuracy through a sophisticated blend of sentiment analysis and ML but also introduces novel methodologies in fake tweet detection, feature extraction, and prediction models, charting new territories in the pursuit of precision and reliability. Results reveal that the proposed method performs better in terms of overall accuracy than the most recent methods. We intend to expand the functionality of our approach in the future by including additional well-known social media, such as Facebook and Google News, etc.