Stock Market Predictions Using Machine Learning Techniques

Abstract

Accurately predicting stock market prices is vital in today’s economy, leading researchers to explore novel approaches for forecasting. Recent studies have shown that historical stock data, search engine queries, and social mood from platforms like Twitter and news websites can predict future stock prices. Previous research often lacked comprehensive data, especially concerning social mood. This study presents an effective method to integrate multiple information sources to address this gap and enhance prediction accuracy. We utilized Long Short-Term Memory (LSTM) and Recurrent Neural Network (RNN) models to analyse individual data sources. To further improve prediction accuracy, we employed an ensemble method combining Weighted Average and Differential Evolution techniques. The results yielded precise forecasts for one-day, seven-day, 15-day, and 30- day intervals, providing valuable insights for investors and helping companies gauge their future market performance. Keywords-- Stock market prediction; Sentiment Analysis; Neural Networks; Long-short Term Memory Neural Networks, DJIA, Ensemble Method, Weighted Average