A Systematic Deep Learning Approach to Forecast Nifty50 Index Trend

The fluctuations in stock prices are highly volatile and unpredictable which obstructs the retail investors from making profitable investment decisions. In this context, a deep learning approach has been devised to capture the trend movement of Nifty50 stocks. The data from various social media platforms, including Twitter, Facebook, StockTwits, and YouTube have been gathered using multiple web scraping tools. The collected data was cleaned and pre-processed to find the subjectivity and polarity scores. These scores were used to determine positive, neutral, and negative sentiments. A novel approach of combining the sentiment scores and market positions was then adopted to predict the Nifty50 returns using the deep learning model. The proposed model comprised long short-term memory networks and densely connected neural networks. Carefully selecting model hyperparameters helped achieve more than 96% prediction accuracies. The prediction accuracy harnessed the investors to strengthen their stock's predicting needs.