Integrating deep learning and optimization algorithms to forecast real-time stock prices for intraday traders

Abstract

The number of stock investors is steadily increasing due to factors such as the availability of high-speed internet, smart trading platforms, lower trading commissions, and the perception that trading is an effective way of earning extra income to enhance financial stability. Accurate forecasting is crucial to earning profits in the stock market, as it allows traders to anticipate price changes and make strategic investments. The traders must skillfully negotiate short-term market changes to maximize gains and minimize losses, as intraday profit mostly depends on the timing of buy and sell decisions. In the presented work, we provide minute-by-minute forecasts that assist intraday traders in making the best decisions on when to buy and sell, consequently maximizing profits on each trade they make. We have implemented a one-dimensional convolutional neural network and bidirectional long-short-term memory (1DCNN-BiLSTM) optimized with particle swarm optimizer (PSO) to forecast the value of stocks for each minute using real-time data extracted from Yahoo Finance. The proposed method is evaluated against state-of-the-art technology, and the results demonstrate its strong potential to accurately forecast the opening price, stock movement, and price for the next timeframe. This provides valuable insights for intraday traders to make informed buy or sell decisions.