Stock Market Prediction System Using LSTM with Technical Indicators as Voters

The stock market is a current significant trend in modern life who seek to improve their wealth through investing, thus accurate stock market predictions are essential to maximize gains and minimize losses. This research focuses on enhancing stock market prediction performance using a combination of several technical indicators and artificial neural network technology, specifically Long Short-Term Memory (LSTM). Technical indicators used include Moving Average Convergence Divergence (MACD), Bollinger Bands, Relative Volume (RVOL), Williams %R, Chaikin Money Flow (CMF), Detrended Price Oscillator, Stochastic Oscillator, Volume Weighted Average Price (VWAP), Relative Strength Index (RSI), Moving Average (MA), and TRIX. Each technical indicator generates buy or sell signals that collectively determine whether a stock is worth buying, selling, or holding. Historical stock price data is collected and preprocessed, followed by training and testing the LSTM model with accuracy metrics to evaluate the performance. Using appl dataset from Kaggle, the training and testing generated RMSE values of 0.004 and 0.04 respectively, whereas the standard deviation is 0.04 and 0.16. These results demonstrate high accuracy which corresponds to our expectations given that a threshold of 5 or below represents the minority and a threshold of 6 or above enters the majority voting region. In the future, we plan to incorporate more variables into the LSTM model to achieve greater accuracy and use different stocks to test our findings.