Stock Forecasting Using Natural Language and Recurrent Network

The factors affecting the stock price of a company include financial aspects, government policies, international policies, emerging news, which affects the investors and hence the market. Critical technical indicators are taken into account while proposing multi-level machine learning in this paper. Firstly, the sentiments of news articles analyzed using lexicon-based Natural Language Processing (NLP). The lexicon used is exclusively based on financial, social media. Each news article analyzed for aspect extraction and aspect-based sentiment analysis. Along with market news sentiments and the company’s historical financial data, feature vector also includes critical technical analyses based on trend, momentum, and volatility. The future stock price movements forecast utilizes Long Short-Term Memory-Recursive Neural Network (LSTM-RNN) model. The results indicate that the discussed model performs well without requiring any data preprocessing, cycle analyses, or seasonality testing.