Stock Index Prediction with Financial News Sentiments and Technical Indicators

Abstract

The price of a stock in the market can be influenced by many factors, out of which the sentiment of the investors plays a vital role. Most often, the sentiment of investors depends on the sentiment of the news headlines. Therefore, news headlines also play an important role in the fluctuation of the stock index. This paper uses the combination of Bidirectional Encoder Representations from Transformers (BERT) and Bidirectional Gated Recurrent Unit (BiGRU) algorithms for the prediction of news sentiment scores based on national news headlines and financial news data. Technical indicators like Relative Strength Index (RSI), Exponential Moving Average (EMA), Moving Average Convergence Divergence (MACD), Stochastic Oscillator along with normal stock indicators like ’Date’, ’Open’, ’Close’, ’High’, ’Low’ and ’Volume’ data can be used to predict the short-term momentum of the stock value. This paper uses the BiGRU algorithm to predict the stock index value (a) with technical indicators only and (b) with technical indicators and news sentiment scores. Keeping all the hyperparameters constant, the BiGRU algorithm provided better prediction results when news sentiment scores were added to the dataset along with technical parameters as an input.